Book a call

A Fable InLead playbook

The complete AEO playbook. Get your brand into AI answers.

42 plays you can run with free tools. For ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, Copilot and Claude.

By Saiful Islam Saif, Founder, Fable InLead. . 45-minute read.

Based on the Ahrefs Academy AEO course by Sam Oh (12 lessons). Written in our own words, with our own data, tools and templates added. Every number links to its dated source.

Fable InLead flame mark
0.664

Correlation between web mentions of a brand and its AI Overview visibility. S12

28%

Of Ahrefs’ own AI mentions carried a link. The rest were names only. S11

14%

Overlap in the top 50 sources named by three AI engines. S7

23x

Higher conversion from AI visitors than organic visitors, at Ahrefs. S2

Get the PDF.

The whole playbook as a 43-page PDF, free. Share your work email and download it at once.

We use your details to give you the download. See our privacy policy.

What this playbook gives you.

42 plays you can run with free tools. Each one has the steps, the tool and the time. It also tells you when you are done.

0.664

Correlation between web mentions of a brand and its AI Overview visibility (S12).

28%

Of Ahrefs’ own AI mentions carried a link. The rest were names only (S11).

14%

Overlap in the top 50 sources named by three AI engines (S7).

23x

Higher conversion from AI visitors than organic visitors, at Ahrefs (S2).

Where it comes from.

Based on the Ahrefs Academy AEO course by Sam Oh. All 12 lessons, read on 17 October 2026. Our own words throughout.

What we added.

Free-tool steps, our own fableinlead.com baseline, six templates, a 30/60/90-day plan and a 20-minute audit.

How to read the tags.

L1.1 means course lesson 1.1. S5 means source 5 in the list at the back. Times are our estimates.

Contents.

Read Part 1 first. Then run the plays in order, or jump to the one you need. Lesson tags show which Ahrefs lesson each chapter covers.

  1. Part 1. How AI answers work
  2. 01What AEO is, and why nowL0.14
  3. 02How AI engines pick their sourcesL1.15
  4. 03Every engine plays by its own rulesL1.26
  5. 04The three kinds of AI visibilityL1.37
  6. 05The 20-minute self-auditours9
  7. Part 2. Strategy
  8. 06Run a brand gap analysisL2.111
  9. 07Keyword and prompt researchL2.214
  10. Part 3. Execution
  11. 08Write pages AI wants to citeL3.118
  12. 09Earn mentions where AI readsL3.220
  13. 10YouTube for AI answersL3.323
  14. 11Technical checksL3.424
  15. Part 4. Measurement
  16. 12Track AI traffic, bots and leadsL4.128
  17. 13Monthly checks, business case and false factsL4.230
  18. Part 5. Do it
  19. 14The free tool kit, step by stepours33
  20. 15Worked example: fableinlead.comours35
  21. 16Your 30/60/90-day planours37
  22. 17Six templatesours38
  23. 18Common mistakesall40
  24. Reference
  25. 19Sources and credits41

Where it comes from.

Based on the Ahrefs Academy AEO course by Sam Oh. All 12 lessons, read on 17 October 2026. Our own words throughout.

What we added.

Free-tool steps, our own fableinlead.com baseline, six templates, a 30/60/90-day plan and a 20-minute audit.

How to read the tags.

L1.1 means course lesson 1.1. S5 means source 5 in the list at the back. Times are our estimates.

Part 1

How AI answers work.

Learn the rules before you play. Four short chapters on how the engines pick names. Then a 20-minute audit of your own brand.

  1. 01What AEO is, and why nowL0.1
  2. 02How AI engines pick their sourcesL1.1
  3. 03Every engine plays by its own rulesL1.2
  4. 04The three kinds of AI visibilityL1.3
  5. 05The 20-minute self-auditours
Chapter 01

What AEO is, and why now.

AEO means answer engine optimization. It is the work of getting your brand named when someone asks an AI tool. Ideally with a link.

Covers Ahrefs lesson 0.1, What is AEO?

Classic SEO fights for a slot in a list of ten blue links. AEO has no list. The AI reads many pages, writes one answer, and decides which brands to mention. L0.1

People also call it GEO (generative engine optimization) or AI SEO. They mean the same work.

Top Google result
58%

Lower click rate when an AI Overview shows, as of December 2025. S1

Ahrefs, June 2025
0.5%

Of site visits came from AI search. Those visits drove 12.1% of sign-ups. S2

AI referral traffic
9.7x

Growth in one year, per an Ahrefs study published June 2025. S3

What the numbers mean for you

AI traffic is still a thin slice of most sites. But it converts well. At Ahrefs, 0.5% of visits brought 12.1% of sign-ups. Ahrefs puts that at 23 times the conversion rate of organic search. S2 The course gives one likely reason. The AI has already told the buyer why you fit. L4.2

That is one company’s data. Measure your own before you plan on it.

AEO sits on top of SEO

The course is clear on this. Good SEO is the base, and AEO is the layer you add. L0.1 The lesson cites Ethan Smith of Graphite on “zero-sum bias”. That is the habit of assuming a new channel must kill the old one.

Ahrefs checked its own data. Its content and product pages had been mentioned 7,470 times across 2,309 pages in AI search. That came with no effort to optimize for it. Years of solid SEO and useful content did the job. L0.1 S4

So if your SEO is weak, start there. Everything in this playbook builds on it.

Chapter 02

How AI engines pick their sources.

Two doors lead into an AI answer. What the model learned in training, and what it finds on the web while it answers.

Covers Ahrefs lesson 1.1, How AI Search Engines Work

Door 1

Training data.

A fixed snapshot of the web: sites, books, PDFs, social posts, video transcripts. The lesson says it refreshes about every six months. To get in, be mentioned widely and often, over time. L1.1

Door 2

Live retrieval.

The AI runs web searches, reads the pages that come back, then writes. The field calls this RAG (retrieval-augmented generation). To get in, rank for the searches it runs. That part is classic SEO. L1.1

One prompt becomes many searches

The AI does not search for the exact words typed. It splits one prompt into many smaller searches and runs them at once. This is called query fan-out. L1.1

One question, many searches, one answer.
A made-up example of query fan-out for a B2B buyer.
PROMPT Which agency can book B2B meetings for us in Singapore? FAN-OUT SEARCHES appointment setting Singapore lead generation agency reviews B2B cost per meeting 2026 best outbound agencies Asia LinkedIn outreach rules PAGES READ ANSWER Brand Anamed and linked Brand Bnamed, no link Youabsent PROMPT Which agency can book B2B meetings for us in Singapore? FAN-OUT SEARCHES appointment setting Singapore lead generation agency reviews B2B cost per meeting 2026 best outbound agencies Asia LinkedIn outreach rules ANSWER, AFTER READING THE PAGES Brand Anamed and linked Brand Bnamed, no link Youabsent
Our illustration of the process in Ahrefs lesson 1.1. The searches and brands are invented.

Two things follow. First, one page for one keyword is not enough. You need to cover the whole topic, including the narrow sub-questions. Second, do not treat fan-out searches as a keyword list. Ahrefs says over 95% of them have no search volume. They also change from run to run. L1.1

Answers work on odds

Ask the same question five times and your brand might appear three times. That is why the field says “visibility” and not “rank”. L1.1 Always measure over several runs.

Three things raise your odds

Consensus

Many sources, one story.

When many pages say the same thing about you, AI is more likely to repeat it. L1.1

Freshness

About 25% fresher.

AI-cited content is fresher than what ranks in classic search. Ahrefs studied 17 million citations. S5

Authority

38% from the top 10.

Of AI Overview citations, 38% come from pages already in Google’s top 10. S6

Chapter 03

Every engine plays by its own rules.

“AI search” is several channels. Each engine has its own index and its own taste in sources.

Covers Ahrefs lesson 1.2, How Ranking in Google AI Overviews, ChatGPT, and Perplexity are Different

Top 50 sources per engine
14%

Overlap across AI Overviews, ChatGPT and Perplexity. S7 L1.2

On all three engines
7

Domains out of 50 appeared on all three lists. S7

AI Mode vs AI Overviews
13.7%

Citation overlap, though the answers were 86% similar in meaning. S10

EngineWhat it tends to citeWhere to start
Google AI OverviewsEstablished, authoritative sites and Google’s own properties. YouTube is about 5.6% of citations. Reddit shows up a lot. It now pulls more from outside the top 10. L1.2Rank in Google. Add YouTube and Reddit.
Google AI ModeDifferent sources from AI Overviews. YouTube, Google and Wikipedia rank among its top cited domains, with more Facebook and Instagram. L1.2 Cites Quora about 3.5 times and Facebook about twice as much as AI Overviews. S10Video, Q&A sites and social proof.
ChatGPTBig publishers and media, such as Reddit, Wikipedia and Forbes. Partly linked to OpenAI’s deals with publishers. Median Domain Rating of its top cited pages that rank in Google is 90. S8 Overlaps with Google’s top 10 only about 8 to 10% of the time. L1.2Mentions on high-authority editorial sites, Reddit and niche review sites.
PerplexityClosest to Google. About 28.6% of its citations rank in Google’s top 10. S9If you already rank, check here first.
GeminiNot profiled in the course. It can render JavaScript. L3.4 Google says Gemini Apps can ground answers in content from its Search index. The Google-Extended token controls that use. S34 Ahrefs got a link in 16.8% of its own Gemini mentions. S11Rank in Google. Do not block Google-Extended. Track it in your prompt runs.
Microsoft CopilotIt can render JavaScript. L3.4 Bing Webmaster Tools shows how often Copilot and Bing’s AI answers cite your pages. S27Verify your site in Bing Webmaster Tools. Play 38.
ClaudeNot profiled in the course. Its web search appears to use Brave Search. Anthropic had not confirmed it when TechCrunch reported it. S28 Claude passes referral data to your analytics. L4.1Check you show in Brave Search. Our companion Claude playbook goes deeper.

How to pick your engines

  1. Start where the users are. At recording time, Google’s AI features and ChatGPT carried most AI search traffic. Begin there. L1.2
  2. Use what you already have. If you rank in Google, you have a head start on AI Overviews and Perplexity. ChatGPT leans more on editorial mentions. L1.2
  3. Do the shared basics everywhere. Useful content, mentions and topic depth help on every engine. Use platform habits only to set priorities.

One more fact from the lesson. Many of the most cited domains in AI search get almost no Google traffic. So check AI answers directly, even if your Google numbers look fine. L1.2

Chapter 04

The three kinds of AI visibility.

Being named counts, even without a link. Track all three states, per engine.

Covers Ahrefs lesson 1.3, What Is AI Visibility? The 3 Types

State 1

Cited and linked.

The answer links to your page. Direct visits, and the easiest to measure.

State 2

Named, no link.

The answer says your name. Like a friend’s tip to thousands of people. It shows up later as branded search or direct visits.

State 3

Not visible.

Neither. This is the gap to close. You can only fix it if you measure it.

Even Ahrefs gets few links.
Share of Ahrefs’ own brand mentions that carried a link, by engine.
Perplexity51.6%
Google AI Mode36.8%
All engines28%
ChatGPT26.9%
Google AI Overviews10.7%
Source: Ahrefs study of its own brand mentions, published 26 Nov 2025. Quoted in lesson 1.3 (S11). “All engines” is the study average. Your own rates will differ.

Links sit on the busy questions

Weighted by search volume, links show up much more. On Perplexity, links appeared in about 78% of impressions. On Gemini, links were in only 16.8% of mentions, yet 71% of impressions. S11 So links cluster on high-volume questions.

Why a plain mention still counts

Models learn from the web. Each credible page that names you next to a topic is one more example. Ahrefs studied 75,000 brands. Of the factors it tested, branded web mentions had the strongest link to AI Overview visibility. The correlation was 0.664. That beat backlinks, Domain Rating and referring domains. S12

That is a correlation. It shows what visible brands have in common. It does not prove that mentions cause visibility.

What to track

  • Mentions: how often your brand is named.
  • Citations: how often your site is linked as a source.
  • Impressions: how many people could see those answers.
  • AI share of voice: your mentions against your competitors’ mentions. L1.3

Watch the space between impressions and mentions. If AI answers questions on your topic often, but rarely names you, that is your opening. L1.3

What counts as a win depends on your business. L1.3

  • Product brands want a place in “best [category]” answers, where buyers click and buy.
  • Service firms want to be the expert named in step-by-step answers.
  • Publishers get cited for direct facts and build authority, though the click is rare.
  • Video creators can appear as cited YouTube videos.
Chapter 05

The 20-minute self‑audit.

Six checks, a timer and a notepad. Each fail points you to the play that fixes it.

Our own check, built from lessons 1.1, 1.3, 2.2, 3.4 and 4.1

Scoring. Everyone runs Play 2 first, so you have a baseline. Then: five or six passes, jump to Part 3. Three or four, work through Part 2. Two or fewer, fix checks 1 and 2 today, then take the plays in order.

2 MIN

1. Are AI bots allowed in?

Open yourdomain.com/robots.txt. Look for the AI bots listed in Play 25, such as GPTBot, ClaudeBot and PerplexityBot. A “Disallow: /“ under any of them is a fail. So is “Disallow: /“ under “User-agent: *”. If you use Cloudflare, check its AI bot setting too. Fail? Go to Play 25.

3 MIN

2. Can a bot read your page without JavaScript?

In Chrome, open DevTools, press Cmd+Shift+P (Ctrl+Shift+P on Windows), type “Disable JavaScript” and reload your homepage. Main text gone? That is a fail, because ChatGPT’s crawler does not run JavaScript. L3.4 Go to Play 26.

5 MIN

3. What does AI say about you?

Ask “What is [your brand]?” in ChatGPT, Perplexity and Gemini. Use a temporary or private chat. Note: right, wrong or “I don’t know”. Wrong or unknown is a fail. Go to Plays 1 and 35.

5 MIN

4. Are you named when a buyer asks?

Ask one real buying question. For example: “Best [your service] for [your buyer] in [your market]?” Use ChatGPT, Perplexity and Google AI Mode. Note who is named and which pages are cited. Not named anywhere? Go to Plays 2 and 10.

2 MIN

5. Is Google already showing you in AI answers?

In Search Console, open Performance and look for the generative AI report. Google had rolled it out to all sites by 31 Aug 2026. S26 Zero AI impressions on your key pages? Go to Play 37.

3 MIN

6. Are you on the lists AI reads?

Search Google for intitle:best "[your category]" -"[your brand]". Count the list pages that name rivals but not you. Three or more is a fail. Go to Play 18.

The full playbook. 42 plays, six templates, one plan.
Part 2

Strategy.

Know where you stand before you build. A gap analysis shows where you are missing. Keyword and prompt research shows where to aim.

  1. 06Run a brand gap analysisL2.1
  2. 07Keyword and prompt researchL2.2
Chapter 06

Run a brand gap analysis.

Where should you appear, and where do you appear today? The gap between the two is your plan.

Covers Ahrefs lesson 2.1, How to Run a Brand Gap Analysis. It draws on Despina Gavoyannis’s Ahrefs guide. S13

Play 01

Map your brand entities.

Engines do not know a name by itself. They learn it from how it is described. Each entity has its own visibility. L2.1

  1. List every entity. That means your main brand, sub-brands and products. Add your own methods and metrics. Add the people tied to the brand, such as the founder.
  2. For each one, write the topics it should be known for, in one line.
  3. Add the words buyers use in your category, such as “boutique”, “enterprise grade” or “affordable”. Pull them from keyword tools and from your own sales calls.
  4. Note where each entity is described today (site, LinkedIn, directories) and whether the wording matches.
  5. Write one short brand definition. Use the same words everywhere.
Free tool
A spreadsheet, plus Template C in chapter 17
Time
45 minutes
Done when
Every entity has topics, buyer words and one agreed definition
Play 02

Take your baseline in AI answers.

You cannot close a gap you have not measured. The course uses Ahrefs Brand Radar, a paid tool. A manual prompt log gives you three of its four views for free. Those are mentions, citations and share of voice. For impressions, use Search Console’s AI report (Play 37). L2.1

  1. Write 10 to 20 prompts the way buyers ask. Mix buying questions, comparisons and costs. Add “What is [brand]?” and “Who is [founder]?”.
  2. Run each prompt in a fresh chat in each engine you care about. Use temporary, incognito or signed-out mode, with memory off.
  3. For each run, record four things:
    • named, linked or absent
    • what it said about you, and whether it is right
    • every competitor named
    • every source domain shown
  4. Answers vary from run to run. L1.1 So run your 3 to 5 key prompts three times each. Report “named in 2 of 3 runs”, never a single result.
  5. Score it: mentions, citations, and share of voice. Share of voice is your mentions divided by all tracked brand mentions in the same runs.
  6. Save the prompt wording. Never edit it between months, or the trend breaks.
  7. Add an SEO snapshot from Play 39: Domain Rating, referring domains, organic keywords and traffic. AI visibility builds on these.
Free tool
The engines, Template A, and Semrush’s or Ahrefs’ free AI checker to cross-check (Plays 39, 40)
Time
2 to 3 hours for 10 prompts in 5 engines
Done when
A dated prompt log and a score per engine
Play 03

Sort what you find into six gaps.

Most of your gaps will fall into one or two of these. That tells you where the plan’s weight goes. L2.1

GapQuestion to askWhere to look
VisibilityDo rivals show up more often than you?Share of voice in your prompt log
NarrativeDoes AI describe you the way you want?“What is [brand]?” answers
TopicAre you missing from topics you should own?Buying prompts where you are absent
FormatDo rivals have cited videos, guides or comparisons you lack?Source domains and page types in the log
Web mentionsDo lists, reviews and forums name rivals but not you?Cited pages, plus Play 18 searches
DemandDo buyers search your space without your name coming up?Search Console queries, keyword tools

One quick win to pull out now. Rows where you were named but not linked are missed citations. Add each to your Fix list. Then make sure the page that should be cited answers that question in its first lines. L2.1

Free tool
Template B in chapter 17
Time
1 hour
Done when
Every gap has dated evidence next to it
Play 04

Fix, build or influence. Quick wins first.

You cannot close every gap at once. Each gap gets one of three actions. L2.1

Fix

Improve what exists.

A ranking page missing a subtopic. Low effort, often high impact.

Build

Make something new.

A page or asset for a need you do not cover at all.

Influence

Work off your site.

Outreach and mentions, for example a big list that names every rival but you.

  1. Tag each gap Fix, Build or Influence.
  2. Give each an effort (S, M, L) and a likely impact (S, M, L).
  3. Do the low-effort, high-impact rows first. Name one owner and one date per row.
Free tool
Template B
Time
30 minutes
Done when
A ranked list with owners and dates
Play 05

Run the same audit on two rivals.

Their branded searches show what buyers and AI tie to them. Often you have the same feature but no page that links it to your name. L2.1

  1. Pick the two rivals named most in your prompt log.
  2. Type each rival’s name into Google and read the autocomplete. Note phrases like “[rival] pricing”, “[rival] vs” and “[rival] alternatives”. These are their branded searches.
  3. Ask each engine “What is [rival]?” and note the traits it repeats.
  4. Open the pages cited for them. Note the page type: market page, cost page, list, comparison, video.
  5. Do they own a trait you share? Add a Fix or Build row to your gap table.
Free tool
The engines, Google, Ubersuggest’s competitor view (free plan limits apply)
Time
1 hour
Done when
2 to 5 new rows in the gap table from rival pages
Chapter 07

Keyword and prompt research.

Most of keyword research is unchanged. One extra test is new. Does the AI answer so fully that nobody clicks?

Covers Ahrefs lesson 2.2, Keyword and Prompt Research for AI SEO

Play 06

Build seeds and modifiers.

Seeds are broad one- or two-word terms. Modifiers like “best” or “how to” turn them into real searches. An AI assistant can draft both. L2.2

  1. Ask any assistant: “We are [business], selling [offer] to [buyers] in [market]. List 10 short topics our buyers search for. Then list words they add, like best, cost, vs, template or near me.”
  2. Put the seeds into a keyword tool. In Semrush free, use Keyword Magic Tool, then “Include keywords” for your modifiers. In Ubersuggest, use Keyword Ideas. Google Keyword Planner is a free fallback with a Google Ads account.
  3. Cross-check volumes in a second tool. When they disagree, plan on the lower number.
Free tool
Any assistant, Ubersuggest and Semrush free (daily limits), Search Console queries
Time
45 minutes
Done when
A list of real buyer phrases with two volume figures
Play 07

Vet every keyword with the BID test.

Some keywords look good and earn nothing. Run three tests before you commit. L2.2

TestAskHow
B, Business potentialWould ranking first help sales?Prefer phrases with intent or budget over plain definitions.
I, IntentDoes your page type match what ranks?Google it. If the top results are a different page type, match it or move on.
D, DifficultyCan you realistically rank?Check links and authority of the top pages. A few weak sites in the top 10 is a good sign.
Free tool
Google, plus Ahrefs’ free backlink checker or Ubersuggest for top-page links
Time
3 minutes a keyword
Done when
Each keyword marked pass or fail on B, I and D
Play 08

Apply the AI filter.

If the AI answer fully satisfies the searcher, nobody clicks. The keyword stays on your list, but its goal changes from a click to a mention. L2.2

  1. Search each keeper keyword in a private window.
  2. Read the AI Overview as the buyer would. Ask: would I still need to click?
  3. If yes, keep it as a click target. If no, move it to your mention list (Play 10).
Questions trigger AI Overviews most.
Share of keywords that show an AI Overview.
Question queries~58%
Queries of 7+ words46%
All keywords~21%
Source: Ahrefs study of what triggers AI Overviews, published 10 Nov 2025. Quoted in lesson 2.2 (S14). In it, 99.9% of trigger keywords were informational.
Free tool
Google in a private window
Time
1 minute a keyword
Done when
Every keeper is tagged “click” or “mention”
Play 09

Find free-tool searches.

Tool searches rarely show an AI Overview, because the searcher needs to use something. They are still good click targets. L2.2

  1. Add these modifiers to your seeds: calculator, checker, generator, tool, template, finder, planner, maker.
  2. Pick 2 or 3 with business potential. A service firm might build a cost calculator or a free template.
  3. Check each in Google for an AI Overview before you build.
Free tool
Ubersuggest or Semrush free, Google
Time
30 minutes
Done when
One tool idea with volume and no AI Overview
Play 10

Find the questions where rivals are named and you are not.

“Best X” lists are the most cited page type. Ahrefs studied ChatGPT’s sources for 750 “best” prompts. Lists made up 43.8% of cited pages. S15 Several lists naming one brand look like consensus. L2.2

  1. From your prompt log, filter rows where you are absent and a rival is named.
  2. Keep the ones with “best”, “top”, “vs”, “review” or “alternative” in them. This is your short list.
  3. Google each phrase. Note which list and comparison pages rank. These feed Play 18.
  4. Re-check monthly. Over 45% of AI Overview citations changed on refresh, and refreshes came about every 2 days. S16
Free tool
Your prompt log, Google
Time
1 hour
Done when
A short list of 10 to 20 mention targets
Play 11

Treat prompts as a sample of the topic.

The same need, asked ten ways, can name ten different sets of brands. So you cannot chase prompts one by one. L2.2

  1. Keep your fixed prompt set for measuring only.
  2. For each prompt, write the topic behind it in three words.
  3. Look for the sub-searches the AI ran. Paid tools such as Ahrefs Brand Radar show them for ChatGPT and Perplexity. Treat them as clues to the sub-topics the AI cares about, never as keywords. L1.1
  4. Build pages, mentions and videos for the topic, never for the exact sentence.
Free tool
Template A
Time
20 minutes
Done when
Every tracked prompt maps to a topic and at least one planned asset
Part 3

Execution.

Four fronts. The pages you write and the mentions you earn. The videos you make, and the technical doors you keep open.

  1. 08Write pages AI wants to citeL3.1
  2. 09Earn mentions where AI readsL3.2
  3. 10YouTube for AI answersL3.3
  4. 11Technical checksL3.4
Chapter 08

Write pages AI wants to cite.

There is no special AI format. Write clearly for people. A few habits help people and machines at once.

Covers Ahrefs lesson 3.1, How to Optimize Content for AI Search Engines

Word count vs being cited
0.04

Correlation, in a study of 174,000+ pages cited in AI Overviews. Length barely matters. S17

Cited pages
53.4%

Were under 1,000 words in the same study. S17

ChatGPT’s top cited pages
76%

Were refreshed in the last 30 days, of pages with a detectable update date. Of those, 89.7% were updated in 2025. S8 L3.1

Play 12

Put the answer first.

Readers and models both pay most attention to the start. The course calls this BLUF: bottom line up front. L3.1

  1. Open your key page. Read the first sentence of the page and of each section.
  2. If it starts with background, move the answer up. Open with the plain answer, such as “Yes, you can, and here is how.” Put the history after it.
  3. Keep the detail below the answer.
Free tool
Your CMS
Time
15 minutes a page
Done when
Every section opens with its answer
Play 13

Make every section stand alone.

AI splits pages into chunks, and models do it in different ways. You do not control where the cuts fall. A section that leans on earlier context loses its meaning. L3.1 S18

  1. Copy one H2 section into a blank document.
  2. Read it cold. Does it make sense alone?
  3. If not, replace “this”, “it” and “as above” with the real names. Repeat for every H2.
Free tool
A blank document
Time
2 minutes a section
Done when
Every H2 passes the cold read
Play 14

Name real things, in short sentences.

AI reads text through entities: brands, products, people, places, methods. Short sentences with one idea each are easiest to parse. L3.1

  1. Highlight every vague phrase: “our solution”, “the platform”, “many markets”.
  2. Replace each with the real name: the service, the country, the buyer role, the method.
  3. Split any sentence you had to read twice. Aim for subject, verb, object.
Free tool
Hemingway Editor (free web version) or a read-aloud
Time
20 minutes a page
Done when
No vague nouns left; no sentence needs a second read
Play 15

Put your brand name on your ideas.

Models tend to absorb an original idea and drop the credit. Naming it with your brand keeps the link. L3.1

  1. Pick one method, framework or metric that you created.
  2. Name it with your brand: “[Brand] [method name]”.
  3. Define it in one plain paragraph on one page of your site.
  4. Use the same name and definition on your blog, social posts, podcasts and forum answers.
Free tool
Your site and channels
Time
2 hours to define, then ongoing
Done when
The named method appears on 3+ places with the same definition
Play 16

Refresh your sleeper pages.

Sleeper pages used to do well and slipped. They already have links. A real update can be one of the fastest wins, because freshness counts. L3.1

  1. In Search Console, open Performance, then Search results. Click Date, then Compare, and pick the last 3 months against the 3 before.
  2. Open the Pages tab and sort by clicks difference. Note pages with big drops.
  3. Check each page in Ahrefs Webmaster Tools, Site Explorer. Look at lost referring domains for the same period. If many links were lost, it is a links problem, so do not rewrite. If links held, the content is stale.
  4. Update for real: new facts, new examples, current numbers with dates. A new date on old text does not count, and Google can tell. L3.1
  5. Show a visible “updated” date that matches a real change.
Free tool
Search Console, Ahrefs Webmaster Tools
Time
1 to 2 hours a page
Done when
Your top 5 to 10 sleepers carry real, dated updates
Play 17

Pick the format AI cites.

Lists, comparisons and original data get cited most. “X vs Y” pages match how people ask. L3.1 S15

  1. For each topic on your list, check the page types cited in your prompt log.
  2. Match the winning format: a “best” list, comparison, cost guide, data page or how-to.
  3. If you have your own numbers, publish them with method notes and a date. AI likes citing specific numbers.
  4. Brief the page with Template D before anyone writes.
Free tool
Prompt log, Template D
Time
30 minutes a brief
Done when
Each planned page has a format that matches what gets cited
Chapter 09

Earn mentions where AI reads.

Much of AEO happens on other people’s pages. Where you are named matters as much as how often.

Covers Ahrefs lesson 3.2, Brand Mentions: How to Get Cited by AI

Branded web mentions
0.664

Correlation with AI Overview visibility across 75,000 brands. Backlinks scored 0.218. S12

Mentions on highly linked pages
~0.7

Correlation with AI Overview visibility, per lesson 3.2. Quality of the page matters. L3.2

ChatGPT sources on “best” prompts
43.8%

Were “best X” lists. Being on them puts you in the pool. S15

Tier 1

Editorial.

Industry publications, review sites, lists and comparisons, YouTube reviews. Hardest to earn, highest value.

Tier 2

Community.

Reddit, Quora, niche forums. Reddit is among the sources ChatGPT cites most. S19

Tier 3

Your own channels.

Other sites you own, your YouTube, podcast and LinkedIn. All are indexed and can be cited.

Play 18

Build your Tier 1 target list.

Get onto the pages AI already cites. Then get onto well-linked lists it may cite next. L3.2

  1. From your prompt log, tally every source domain per engine. Keep lists, reviews and directories. Skip rivals’ own pages.
  2. Find future candidates in Google: intitle:best "[topic]" -"[brand]", intitle:vs "[topic]" and intitle:top "[topic]" [country].
  3. Check each page’s links in Ahrefs’ free backlink checker. Prefer pages with many referring domains.
  4. Find the author or editor. Check the byline, their LinkedIn, and any “submit a tool” or “write for us” page. A free email finder such as Hunter can help. Send a short, honest note with real facts about you. No fake reviews and no hidden paid placements.
  5. Keep the note to three lines. A pattern you can adapt:
    Hi [name], your list "[title]" ranks for [phrase].
    We are [one-line definition], serving [buyer] in [market].
    Happy to send facts if you plan a refresh.
  6. Log each in Template E with a status and date.
Free tool
Prompt log, Google operators, Ahrefs free backlink checker
Time
3 hours for the first 10
Done when
10 targets with contact routes and a first note sent
Play 19

Help in the threads AI reads.

Spam in threads backfires fast and hurts your name. Answer real questions where your experience helps the person asking. L3.2

  1. Note Reddit and Quora threads cited in your prompt runs.
  2. Google your buyer phrases with no operator. Note any Reddit or Quora thread in the top 5 results. Also try [phrase] reddit. Repeat for niche forums.
  3. Read each subreddit’s rules first.
  4. Answer in three parts. First the direct answer. Then one real example from your work. Then say who you are, so nobody feels tricked. Link only if the rules allow it and the link answers the question.
  5. Aim for two or three full answers a month.
Free tool
Google, Reddit, Quora
Time
30 minutes an answer
Done when
3 helpful answers live in threads that rank
Play 20

Use your own channels as sources.

Your LinkedIn, YouTube and podcast are indexed. They can be cited too. L3.2

  1. Use your one brand definition (Play 1) in every profile and bio.
  2. Publish your named method (Play 15) on LinkedIn as a post or article. Use the same words.
  3. Link profiles to your site and back, so they read as one entity.
Free tool
LinkedIn, YouTube, your site
Time
1 hour, then ongoing
Done when
Every profile uses the same definition and links home
Play 21

Audit your mentions every month.

Mentions vanish when lists are updated. AI can also pick up wrong facts. The longer a wrong fact stays up, the more AI learns from it. L3.2

  1. Re-check every page in Template E. Still names you? Still correct?
  2. If a page dropped you after an update, ask the editor why.
  3. Read what each engine says about you. Note anything wrong or unfair.
  4. For a wrong fact, correct it on your own site first. Then ask the source to fix theirs.
Free tool
Template E, Google Alerts for your brand name
Time
45 minutes a month
Done when
Every tracked mention checked and every wrong fact actioned
Chapter 10

YouTube for AI answers.

YouTube is the most cited domain in Google’s AI Overviews, per the course. L3.3 Ahrefs studied 75,000 brands. YouTube mentions had the strongest correlation with visibility, about 0.737. That held across ChatGPT, AI Mode and AI Overviews. S20

Covers Ahrefs lesson 3.3, YouTube SEO for AI Search

Play 22

Pick search hits, not viral hits.

A search hit gets steady views every month because people look for the topic. Videos Google already ranks are likely AI Overview picks. L3.3

  1. Google your niche phrases. Note which ones show videos on page one.
  2. Check each phrase’s volume in Ubersuggest or Semrush free.
  3. Keep 3 to 5 phrases where buyers search and videos rank.
Free tool
Google, Ubersuggest or Semrush free
Time
40 minutes
Done when
3 to 5 video topics with search volume
Play 23

Make the video rank.

Five set-up steps decide whether Google can match the video to the search. L3.3

  1. Put the exact search phrase in the title. Use the thumbnail for the eye-catching part.
  2. Write a real summary in the description, with the phrase in the first lines.
  3. Add timestamps, so YouTube makes chapters that Google can show for sub-questions.
  4. Say the phrase out loud in the video. The lesson cites Google’s Liz Reid confirming Google understands the audio.
  5. Match the format that already ranks: tutorial, list or review.
Free tool
YouTube Studio
Time
20 minutes of set-up a video
Done when
All five boxes ticked before you publish
Play 24

Add the AI layer.

Find the topics where AI answers cite YouTube, then fill them. Every video is also possible training data. L3.3

  1. In your prompt log, mark each run that cites a YouTube link.
  2. List those prompts’ topics. Those are your best video topics.
  3. Add one search-led video a month on a buyer question your site already answers.
Free tool
Prompt log
Time
15 minutes a month
Done when
A list of topics where AI already cites video
Chapter 11

Technical checks.

A blocked page cannot be cited. Technical AEO means making sure AI can reach and read your pages. No rebuild needed.

Covers Ahrefs lesson 3.4, Technical SEO for AI: Robots.txt, GPTBot and llms.txt

Play 25

Open the doors: robots.txt and your CDN.

About 5.9% of 140 million sites block GPTBot, per Ahrefs. S21 Many do it by accident. L3.4

  1. Open yourdomain.com/robots.txt. Search for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot and Google-Extended.
  2. Remove any Disallow: / under the bots you want in. Example of an open rule:
    User-agent: GPTBot
    Allow: /
    
    User-agent: OAI-SearchBot
    Allow: /
  3. Check your CDN too. Cloudflare changed its default to block AI crawlers on 1 July 2025. S35 Since 15 September 2026, new domains block training and agent bots on pages with ads. Search bots stay allowed. S38 Cloudflare counts chat fetch bots, such as ChatGPT-User and Claude-User, as agent bots. Check that setting too. In Cloudflare, check Security settings, then the AI bot policies, and the AI Crawl Control page. S36 A CDN block overrides an open robots.txt.
  4. Know what Google-Extended does. It controls Gemini training and grounding in Gemini Apps. It does not change your place in Google Search. S34 AI Overviews and AI Mode are part of Search, so keep Googlebot open.
  5. Run Ahrefs Webmaster Tools Site Audit. The lesson says Site Audit flags robots rules that block AI bots.
Free tool
Browser, Ahrefs Webmaster Tools, your CDN dashboard
Time
10 minutes
Done when
Every bot you want is allowed in robots.txt and at the CDN
Play 26

Make sure text loads without JavaScript.

Gemini and Copilot can render JavaScript. ChatGPT’s crawler cannot. A page built in the browser looks empty to it. L3.4

  1. Disable JavaScript in your browser and load your key pages (see audit check 2).
  2. Or run curl -s https://yourdomain.com/page | grep "a phrase from your page" in a terminal. No match means the text needs JavaScript.
  3. If text is missing, ask your developer for server-side rendering or static HTML.
Free tool
Chrome DevTools, curl
Time
10 minutes
Done when
Main text of every key page is in the raw HTML
Play 27

Speed up slow pages.

During live retrieval, a slow page can be dropped before it is scored. L3.4

  1. Run each key page through PageSpeed Insights on mobile.
  2. Note Largest Contentful Paint (LCP). Google’s “good” mark is 2.5 seconds or less. S33
  3. Fix the top issues it lists: big images, heavy scripts, slow fonts.
Free tool
PageSpeed Insights, Search Console Core Web Vitals
Time
15 minutes to test; fixes vary
Done when
Key pages pass Core Web Vitals on mobile
Play 28

Clean up headings.

AI may chunk your page at any heading. Logical headings keep each chunk clear. L3.4

  1. One H1 for the page title. H2 for main sections. H3 for sub-sections.
  2. One idea per paragraph.
  3. Check with a free outline extension such as HeadingsMap. Site Audit in Ahrefs Webmaster Tools works too.
Free tool
Ahrefs Webmaster Tools Site Audit
Time
5 minutes a page
Done when
No skipped levels, one H1 per page
Play 29

Decide on schema and llms.txt. Keep both cheap.

Neither has shown a clear effect on citations. Ahrefs checked 137,000 sites. Of those, 28% had an llms.txt file. And 97% of those files got zero requests in May 2026. S22 The course says no major AI provider officially supports llms.txt. It calls the evidence for schema mixed. L3.4 Google says the file is not needed for its AI search features. S22

  1. Keep any schema you already use for SEO (Organization, Article, FAQPage, LocalBusiness).
  2. If you add llms.txt, keep it short and accurate. A file with old facts does more harm than none.
  3. Spend the saved time on Plays 12 to 21.
Free tool
Google Rich Results Test, plus validator.schema.org for other types
Time
20 minutes
Done when
Schema valid; llms.txt facts match your site, or no llms.txt
Play 30

Catch made-up URLs.

AI assistants send visitors to 404 pages 2.87 times as often as Google Search. About 1% of ChatGPT’s clicked URLs hit a 404. S23

  1. Find 404 visits from AI tools. In GA4, open Reports, Engagement, Pages and screens. Filter by your 404 page title. Add Session source as a second dimension. Look for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai. Ahrefs Web Analytics has a ready “Possible 404” report. L3.4
  2. For a made-up URL with steady visits, add a 301 redirect to the closest real page.
  3. Make your 404 page useful: search, top pages, contact.
Free tool
Your analytics, CDN redirect rules, or Ahrefs Web Analytics
Time
15 minutes a month
Done when
No AI-referred 404 with steady visits is left without a redirect
Part 4

Measurement.

No single number tells the whole story. Use three measures together, review monthly, and audit deeply each quarter.

  1. 12Track AI traffic, bots and leadsL4.1
  2. 13Monthly checks, business case and false factsL4.2
Chapter 12

Track AI traffic, bots and leads.

Three measures, each with a blind spot. Together they show the trend.

Covers Ahrefs lesson 4.1, How to Track AI Traffic in GA4 and Ahrefs Web Analytics

Measure 1

AI referral visits.

Which AI tools send people, to which pages. Always an undercount, because some tools strip the referrer.

Measure 2

AI bot activity.

Which pages AI systems fetch. It shows what AI reads. It does not show sales.

Measure 3

Self-reported source.

What buyers say brought them. The best link to revenue, if people answer.

Play 31

Set up an AI traffic channel.

Referral data shows the trend, as a floor. ChatGPT search links pass a referrer; its in-text links for paid users do not. Claude passes it. Perplexity passes it on the web, but not from its desktop app. Copilot passes it on the web, but not in Windows. Grok passes nothing. L4.1 S24

  1. In GA4: Admin, then Data display, then Channel groups. Copy the default group.
  2. Add a channel named “AI traffic”. Set the source to match this regex (chatgpt.com is our addition to the lesson’s list):
    .*(chatgpt\.com|chat\.openai\.com|perplexity|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|deepseek\.com).*
  3. Drag the new channel above “Referral” so it catches these visits first.
  4. View it in Reports, Acquisition, Traffic acquisition, with your new channel group.
  5. Read it monthly. Pages that get AI visits: keep them current, with a clear next step. Key pages with no AI visits: check crawl access and content. L4.1
  6. No GA4? Ahrefs Web Analytics is free and has an AI search channel built in. It also splits unknown traffic from direct, which GA4 does not. L4.1 On Cloudflare Web Analytics, read the referrer list for the same domains each month.
Free tool
GA4, Ahrefs Web Analytics or Cloudflare Web Analytics
Time
20 minutes
Done when
AI visits show as their own channel, by page
Play 32

Watch which pages AI bots fetch.

Training bots collect content. Search and citation bots fetch pages for AI answers. Those drive the referral visits. L4.1 Some index ahead, like OAI-SearchBot. Others fetch live, like ChatGPT-User.

  1. Pick a source. Ahrefs Bot Analytics connects to Cloudflare and works with a free Cloudflare plan. L4.1 Cloudflare’s own AI Crawl Control page is on all plans. S36 Server logs work too.
  2. Split the bots. Training: GPTBot, ClaudeBot. Search and live answers: OAI-SearchBot, Claude-SearchBot, PerplexityBot, ChatGPT-User, Claude-User. Google-Extended is only a robots.txt switch, so it never shows in logs. S34
  3. A live bot hitting one page again and again: that page is likely a source. Keep it fresh.
  4. Key pages never fetched: a discovery problem. Add internal links to them.
Free tool
Cloudflare AI Crawl Control, Ahrefs Bot Analytics, or server logs
Time
30 minutes to connect; 10 minutes a month
Done when
A monthly list of top pages per AI bot
Play 33

Ask “How did you hear about us?”

Much AI influence never shows as a referral. People hear your name in an answer, then type your URL or Google you. At Ahrefs, about 3% of conversions over a year were self-reported as AI. L4.1

  1. Add the question to your sign-up, booking, checkout or post-purchase form.
  2. Options: ChatGPT, Perplexity, Google AI Overviews or AI Mode, Gemini, Copilot, Claude, other AI assistant, Google search, LinkedIn, referral, other.
  3. Review answers monthly next to your referral data.
Free tool
Your form or booking tool
Time
10 minutes
Done when
Every new lead answers the question
Chapter 13

Monthly checks, business case and false facts.

AI traffic is still a thin slice, but it converts well and grows fast. It also builds on the SEO you should do anyway.

Covers Ahrefs lesson 4.2, Your AEO Strategy and Action Plan Checklist

AI referrals today
0.25%

Of a site’s traffic on average, per Ahrefs data quoted in lesson 4.2. L4.2

Google vs AI
210x

More traffic from Google than the top AI platforms combined, same source. L4.2

Self-reported, at Ahrefs
~3%

Of conversions over a year said they came from AI. L4.1

Play 34

Run the monthly check.

Four checks, then log them. Once a month, against your baseline. L4.2

  1. Share of voice. Re-run your fixed prompts. Flat while a rival grows? Find out why.
  2. Cited domains. New sites naming or citing you since last month.
  3. Topic coverage. Gaps closed; new topics where you now appear.
  4. Accuracy and tone. Is AI saying true, fair things?
  5. Add AI referrals, bot fetches and self-reported AI leads. Log it all in Template F.
Free tool
Prompt log, Search Console, Bing Webmaster Tools, analytics
Time
2 hours a month
Done when
One dated row in the monthly report
Play 35

Defend against false facts.

AI can repeat planted lies. Ahrefs tested this with a made-up brand. Gemini and Perplexity repeated false claims in about 37 to 39% of answers. ChatGPT stayed under 7% and cited the brand’s official FAQ in 84% of answers. S25

  1. List every question a buyer might ask about you. Think founding year, founder, location, prices, services and clients.
  2. Answer each one plainly on your own site, with specific numbers and dates. An FAQ page works well.
  3. Use the same facts on every profile and directory.
  4. When AI says something false, publish the correction on your site first. Then ask the outside source to fix theirs.
Free tool
Your site, the engines
Time
2 hours once, then monthly checks
Done when
Every common question has an official answer on your site
Play 36

Run the quarterly deep audit and make the case.

Once a quarter, repeat the full gap analysis against your rivals. Then report in business terms. L4.2

  1. Repeat Plays 2, 3 and 5 with the same prompts.
  2. Check Google’s generative AI report in Search Console for AI impressions by page. S26
  3. Build a before and after table. Real numbers only, each with its tool and date.
  4. Say what moved, what did not, and the next three priorities.
  5. Frame value like brand advertising. Each mention is an impression that may become a branded search.
Free tool
All of chapter 14
Time
Half a day a quarter
Done when
A one-page quarterly report with three next steps
Part 5

Do it.

The course shows paid tools. This part shows how to do the same work for free. Then our own before picture, a plan and the templates.

  1. 14The free tool kit, step by stepours
  2. 15Worked example: fableinlead.comours
  3. 16Your 30/60/90-day planours
  4. 17Six templatesours
  5. 18Common mistakesall
Chapter 14

The free tool kit, step by step.

A handful of free tools cover almost every play. Menus change often, so we date what we saw.

Our own steps, checked 17 October 2026

Play 37

Google Search Console: the AI report.

Google launched generative AI reports on 3 June 2026. All sites had them by 31 August 2026. They show impressions in AI Overviews and AI Mode by page, country, device and date. They do not show clicks. S26

  1. Verify your domain at search.google.com/search-console (a DNS record is best).
  2. Submit your sitemap under Indexing, then Sitemaps.
  3. Open Performance and find the Search generative AI report. Note AI impressions per page.
  4. In the main Search results report, look at queries per page. That is your demand gap data (Play 3).
  5. Use Pages, compare dates, to find sleeper pages (Play 16).
Feeds plays
3, 16, 34, 36
Cost
Free
Watch out
AI data also stays inside your normal totals, so do not add the two together
Play 38

Bing Webmaster Tools: AI Performance.

Microsoft added an AI Performance view on 10 February 2026, as a public preview. It counts how often Copilot, Bing’s AI summaries and some partner tools cite your pages. It shows sample “grounding queries”, the phrases the AI used to find your page. S27

  1. Sign in at bing.com/webmasters. You can import your site from Search Console in one step.
  2. Submit your sitemap.
  3. Open AI Performance. Note total citations and the pages cited. Since June 2026 it also groups grounding queries into intents and topics. It shows your citation share and can compare two periods. S37
  4. Read the grounding queries. They show which sub-topics bring you in.
Feeds plays
2, 11, 34
Cost
Free
Watch out
Citation counts are not ranks, and no click data is shown
Play 39

Ahrefs Webmaster Tools and Semrush free.

Two free accounts give you an SEO snapshot and a site audit. They also give a second opinion on AI mentions.

  1. Ahrefs Webmaster Tools (ahrefs.com/webmaster-tools): verify your site. Run Site Audit for crawl, robots and heading issues (Plays 25, 28).
  2. In the same account, open Site Explorer on your own site. It shows backlinks, organic keywords and top pages (Play 16).
  3. Semrush’s free AI Search Visibility Checker covers ChatGPT, Gemini, AI Mode and AI Overviews. No sign-up. Three uses a day, per its page. S29
  4. Use Semrush’s free daily searches to cross-check keyword volume and difficulty (Play 6).
Feeds plays
2, 6, 7, 16, 25, 28
Cost
Free, with daily limits
Watch out
Tools disagree. Show both figures and plan on the lower one
Play 40

Ahrefs’ free AI checker and Ubersuggest.

Two more free views of your AI presence and keywords.

  1. Ahrefs’ free AI Visibility Checker is at ahrefs.com/ai-visibility-checker. It covers ChatGPT, Gemini, Perplexity, Copilot, AI Overviews and AI Mode. It previews your top 5 mentions by platform. It also lists top topics and cited pages. No sign-up, per its page. S30
  2. Ubersuggest’s free plan gives a few reports a day. We got about three new reports a day in October 2026.
  3. Use it for keyword ideas, a site audit and page speed. Load your fixed prompts into its AI visibility feature if your plan allows.
Feeds plays
2, 5, 6, 9, 22
Cost
Free, with tight daily limits
Watch out
Spread runs over several days
Play 41

Brave Search, for Claude.

Claude’s web search appears to use Brave Search, per TechCrunch on 21 March 2025. Anthropic listed Brave as a subprocessor. It had not confirmed the link when the story ran. S28

  1. Search site:yourdomain.com at search.brave.com. Are your key pages there?
  2. Search your top buyer phrases. Note who ranks.
  3. Missing pages? Brave has a submit page at search.brave.com/submit-url. S31
  4. For a full Claude study, see our companion playbook. It is called “How to get your business mentioned in Claude”.
Feeds plays
2, 34
Cost
Free
Watch out
We found no Brave webmaster console like Google’s or Bing’s (Oct 2026)
Play 42

Manual prompt checks, done fairly.

This is the core measure. It is free and it is what buyers see. It only works if every month runs the same way.

  1. Fixed prompt list, saved, never reworded.
  2. Fresh chat per prompt. Temporary or incognito mode, memory and personalization off. Write down the mode.
  3. Same engines every month. If one needs a login you cannot use, write “not tested” and why.
  4. Record the four things from Play 2. Named, linked or absent; what was said; rivals; sources.
  5. Run your key prompts three times. Save the raw log with the date. Never estimate a result you did not see.
Feeds plays
2, 3, 10, 18, 24, 34
Cost
Free
Watch out
Location and account history can change answers. Note both
Chapter 15

Worked example: fableinlead.com.

Our own before picture. We ran this method on our site in October 2026. Here is what we found, including the weak numbers.

Source: Fable InLead’s own dated logs, 1, 2 and 17 October 2026 (S32)

Fable InLead’s site went live on 30 September 2026. Two days later we ran 10 fixed prompts in five engines. They were ChatGPT, Perplexity, Copilot (Bing Copilot Search, signed out), Gemini and Claude. Fifty answers in all.

Named only when someone typed our name.
Fable InLead in 50 AI answers, 2 October 2026.
Named, all prompts3 of 50
Linked, all prompts1 of 50
Named on buying prompts0 of 40
Most-named rival agency11 of 50
Source: Fable InLead prompt log, 2 October 2026 (S32). Bar length is the share of answers. Rival name withheld.

Our SEO and AI baseline

MeasureWhat we sawTool and date
Google clicks and impressions5 clicks, 188 impressions, average position 29.6, 22 queriesSearch Console, 29 Sep to 14 Oct 2026
Authority scoresAhrefs DR 0, Semrush Authority Score 0, Ubersuggest DA 117 Oct, 16 Oct and Sep 2026
Referring domains1, 40 and 3. Three tools, three answersAhrefs 17 Oct, Semrush 16 Oct, Ubersuggest Sep 2026
AI mentions, second check0 across ChatGPT, AI Overviews, AI Mode and GeminiSemrush free checker, 16 Oct 2026
AI bot accessrobots.txt allows every major AI bot; homepage text loads without JavaScriptcurl check, 17 Oct 2026
Phone speedLCP 3.4 seconds, above the 2.5 second markUbersuggest, logged 17 Oct 2026

What the six gaps looked like

  • Visibility. Rivals were named on almost every buying prompt. We were named in none of the 40 buying answers.
  • Narrative. One engine described our brand backwards: it read our old name as the parent company. Another called us a “platform”. We are a service with a team.
  • Topic. No engine tied us to Southeast Asia or the GCC, our core markets.
  • Format. Rivals won with simple country plus service pages and cost pages. We had few of either.
  • Web mentions. The only outside source the engines had for us was LinkedIn.
  • Demand. Search Console showed 22 queries with impressions. Most of the top 10 were market phrases like “lead generation malaysia”. The engines never paired those markets with our name.

One more gap sits outside the six. All five engines mixed our founder up with other people who share his name. That is an entity problem, so it goes into Play 1.

What we are doing about it

  1. One set of facts everywhere (Plays 1, 35). Same name, definition, founding year and founder line everywhere. That means the site, LinkedIn and every directory, with “formerly Fable Outbound”. We also found our own llms.txt carried two versions of one track-record figure. We are making it one.
  2. Market pages (Plays 10, 17). Our Malaysia market page is the one Google tests most. Singapore and UAE pages follow, built the same way.
  3. Directories and lists (Play 18). Clutch, GoodFirms, Sortlist and DesignRush, plus the “best agencies” roundups the engines cited.
  4. Own the open answers (Plays 15, 17). No agency was named for WhatsApp B2B outreach in the GCC. We have a guide on it.
  5. Measurement (Plays 31 to 33). A “How did you hear about us?” question on our booking form. A monthly check of AI referrals and bots.

Next check: 17 November 2026, same prompts, same engines, plus Google AI Mode and AI Overviews. We will publish the before and after with real numbers only.

Chapter 16

Your 30/60/90-day plan.

Foundation, then build, then compound. One owner and one date on every line.

Built from the first-week plan in lesson 4.2, extended by us

Days 1 to 30

Foundation.

  • Robots.txt and CDN open to AI bots (Play 25)
  • Text loads without JavaScript (Play 26)
  • AI channel, bot report and “How did you hear?” live (Plays 31 to 33)
  • Search Console and Bing Webmaster Tools verified (Plays 37, 38)
  • Entity map and one brand definition (Play 1)
  • Baseline prompt log saved (Plays 2, 42)
  • Wrong facts fixed on your own site (Play 35)
  • Real updates to your top 5 to 10 pages (Play 16)
Days 31 to 60

Build and influence.

  • Gap table sorted into Fix, Build, Influence (Plays 3, 4)
  • 2 to 3 missing pages built: list, comparison, cost or data (Play 17)
  • One named method or data set published (Play 15)
  • Outreach to your top 10 mention targets (Play 18)
  • 3 to 5 helpful forum answers (Play 19)
  • 1 to 2 search-led videos (Plays 22, 23)
  • Month 2 prompt run, same wording (Play 34)
Days 61 to 90

Compound and report.

  • Second refresh of pages AI already sends people to (Play 16)
  • Follow up open outreach; more list placements (Play 18)
  • Mention audit and accuracy check (Play 21)
  • Made-up URLs redirected (Play 30)
  • Quarterly deep audit against rivals (Play 36)
  • Before and after report; next quarter’s three priorities
Chapter 17

Six templates.

Copy them into a spreadsheet. On the web version of this playbook, each one is also a CSV download. Every number gets a source and a date.

Our own templates, built from lessons 1.3, 2.1, 3.1, 3.2 and 4.2

A. Prompt log

One row per run. Header: brand and other names, rivals tracked, run date, run by, mode per engine. N = named, L = linked, A = absent.

Download Template A as a CSV file

Swipe the table sideways to see every column.

#Prompt (exact words)EngineRunResult (N/L/A)What it said (right?)Rivals namedSources shown
1Best [service] for [buyer] in [market]?ChatGPT1ANot mentionedRival A, Rival Brivala.com, a “best agencies” list
1Best [service] for [buyer] in [market]?Perplexity1NNamed third; services right, country wrongRival Aa directory, reddit.com
1Best [service] for [buyer] in [market]?Google AI Mode1ANot mentionedRival Ba “best agencies” list, youtube.com

B. Gap table

Download Template B as a CSV file

Swipe the table sideways to see every column.

GapEvidence (dated)Fix / Build / InfluenceActionOwnerEffortImpact
Visibility
Narrative
Topic
Format
Web mentions
Demand

C. Entity map

Download Template C as a CSV file

Swipe the table sideways to see every column.

EntityTypeShould be known forBuyer wordsDescribed where todaySame wording?
Brand
Founder
Method

D. Page brief

Download Template D as a CSV file

Swipe the table sideways to see every column.

FieldFill in
Page (URL or new)
Question it answers, in buyer words
Prompts in the log it supports
Format: list, comparison, data, how-to, market page
One-line answer for the first sentence
Sections, each able to stand alone
Named entities to include
Original data or named method
Real numbers (source, date)
Internal links in and out
What gets refreshed, and how often

E. Mention tracker

Download Template E as a CSV file

Swipe the table sideways to see every column.

#Site and pageTierAI engines citing itNames rivals?Names you?Contact routeStatusDate
1
2
3

F. Monthly report

One row a month. Under the table, write three lines: what moved, what did not, next three steps.

Download Template F as a CSV file

Swipe the table sideways to see every column.

MonthNamed (runs)Linked (runs)Share of voiceNew citing domainsWrong facts seenAI referralsBot fetchesSelf-reported AIAI impressions (GSC)
Chapter 18

Common mistakes.

Twelve ways to waste a quarter. Each one comes from the course or from our own runs.

  1. Treating AI search as one channel.Each engine cites different sources. Score each one on its own. (L1.2)
  2. Chasing exact prompts or fan-out searches.They change every run and most have no volume. Build for the topic. (L1.1, L2.2)
  3. Reporting one run as a rank.Answers are odds. Report hit rates over repeated runs. (L1.1)
  4. Counting only clicks.In Ahrefs’ own data, about 28% of mentions carried a link. Much of the value shows up later as branded search. (L1.3, S11)
  5. Writing long pages to rank.Length showed almost no link to being cited. Answer the question. (L3.1)
  6. Changing the date without changing the page.Updates must be real. Google can tell. (L3.1)
  7. Blocking AI bots by accident.Check robots.txt and your CDN, and keep text out of JavaScript-only builds. (L3.4, S35)
  8. Expecting llms.txt or schema to do the work.Neither has shown a clear effect on citations. (L3.4)
  9. Spamming Reddit or forums.It backfires fast and hurts your name. Help in real threads. (L3.2)
  10. Leaving gaps in your own facts.AI fills them from weak or false sources. (L4.2)
  11. Rewording your prompts between months.The trend breaks and you cannot compare. (Our runs)
  12. Trusting one SEO tool’s numbers.Three tools gave us 1, 3 and 40 referring domains for one site. Cross-check, then plan on the lowest figure. (Our runs, Sep to Oct 2026)
Chapter 19

Sources and credits.

Based on the Ahrefs Academy AEO course by Sam Oh, VP of Marketing at Ahrefs. 12 lessons, videos published 29 April to 18 May 2026. Lesson pages read in full on 17 October 2026.

A note on weight. Most numbers here come from Ahrefs studies, and Ahrefs sells the tools the course shows. The studies are large and dated, but many are correlations. Treat each one as “Ahrefs found”, with its date. Figures are as quoted in the lessons; some studies have been updated since. AI answers change often.

The 12 lessons

  • L0.1What is AEO? ahrefs.com/academy/aeo-course/lesson-0-1
  • L1.1How AI Search Engines Work. lesson-1-1
  • L1.2How Ranking in Google AI Overviews, ChatGPT, and Perplexity are Different. lesson-1-2
  • L1.3What Is AI Visibility? The 3 Types Every Marketer Needs to Know. lesson-1-3
  • L2.1How to Run a Brand Gap Analysis. lesson-2-1
  • L2.2Keyword & Prompt Research for AI SEO (AEO). lesson-2-2
  • L3.1How to Optimize Content for AI Search Engines. lesson-3-1
  • L3.2Brand Mentions for SEO: How to Get Cited by AI (3 Tiers). lesson-3-2
  • L3.3YouTube SEO for AI Search: How to Rank your YouTube Videos. lesson-3-3
  • L3.4Technical SEO for AI: Robots.txt, GPTBot & llms.txt Explained. lesson-3-4
  • L4.1How to Track AI Traffic in GA4 and Ahrefs Web Analytics (ChatGPT & More). lesson-4-1
  • L4.2Your AEO Strategy + Action Plan Checklist. lesson-4-2

Studies and other sources

Credit. Course ideas are summarized in our own words, with a link to each lesson. No course text is copied. All diagrams and templates are ours. So are Plays 37 to 42, the audit, the plan and the worked example.

Run the audit this week.

Start with chapter 5 and a timer. If you want a second pair of eyes on your results, message Saif on LinkedIn.

linkedin.com/in/saiiful-islam-saif

Honest by design. Boutique by choice.

Fable InLead. fableinlead.comThe complete AEO playbook. Updated 17 October 2026.

Keep a copy.

The same playbook as a 43-page PDF, to read offline or share with your team.

We use your details to give you the download. See our privacy policy.