An AI answer is not a ranking. It is a sentence Google, ChatGPT, or Perplexity writes on your behalf — then links to the page that backs it up. Answer engine optimization (AEO) is the work of making your page the one worth linking: clear enough to quote, trustworthy enough to stand behind, and structured so a machine can lift the right passage.
Google has now written down most of what matters. Its May 2026 guide to optimizing for generative AI features says, from Google Search’s point of view, that optimizing for generative AI search is optimizing for the search experience — still SEO, not a separate discipline. The same guide lists the popular hacks you can ignore. This checklist is built on that documentation, on Google’s helpful-content guidance, and on the best causal evidence about what AI answers do to clicks.
Key Takeaways
- No special markup, file, or sign-up earns a citation. Eligibility is an indexed, snippet-eligible page included in Search generative AI features.
- AI answers are built from passages. A direct answer under a question-form heading is the unit that gets cited.
- Query fan-out means the subtopics around your question matter as much as the question itself — but don’t spin up a separate page for every variation.
- Structured data must match the visible text, and E-E-A-T signals — real authorship, sourcing, first-hand experience — are the trust layer.
- A citation is visibility, not traffic: where an AI Overview appeared, a field experiment found outbound clicks fell 39.8%. Measure Share of Voice, not clicks.
Is Your Page Even Eligible to Be Cited?
A page can only be cited if Google has indexed it and it is eligible to appear with a snippet in search results. Google’s own documentation states there are no additional technical requirements for AI Overviews or AI Mode — no special schema, no sign-up, no separate submission. If your page fails either test, nothing else on this checklist matters.
Check it before touching a word of content. In Search Console, confirm the page is indexed — a stray noindex tag makes citation impossible — and confirm the site is included in Search generative AI features (Settings; included by default). As for opting out: nosnippet stops AI features using your text but also removes your ordinary snippet, while max-snippet and data-nosnippet limit how much is used. Google-Extended does not remove you from AI features in Search.
- ☐ The page is indexed in Search Console, with no noindex tag.
- ☐ The site is included in Search generative AI features in Search Console.
Does Your Page Answer the Question in the First 50 Words?
An AI answer is assembled from passages, and the passage it lifts is usually the one that answers the question outright. Put a direct, self-contained answer in the first paragraph under each question heading — a definition, a number, or a yes-or-no with the condition attached. Detail and examples come after; the answer itself never waits for the third paragraph.
This is the highest-leverage habit in AEO, and the one most pages fail: the typical article opens with three paragraphs of scene-setting before saying anything quotable. Keep the answer honest about its limits: “Yes, if…” beats a bare “yes” the rest of the page contradicts.
- ☐ Every question heading is followed immediately by a direct answer, not an introduction.
- ☐ Each answer paragraph stands alone — definition, figure, or verdict included.
Are Your Headings Written as Questions People Actually Ask?
Google’s AI features use query fan-out: one question becomes several related searches across subtopics before an answer is written. A page whose headings match those sub-questions — the follow-ups, comparisons, and edge cases a real searcher asks next — gives the system more places to cite you. Headings should mirror real questions, not clever labels.
Fan-out is why a page that answers “How much does SEO cost?”, “What affects the price?”, and “Is a monthly retainer worth it?” outperforms a page that only defines the term — the subtopics are where the citations live. Find yours in Search Console queries and People Also Ask boxes. One warning: do not spin up a separate page for every variation; Google says that violates its scaled content abuse policy.
- ☐ Headings are questions in the words searchers use, not branded labels.
- ☐ Each section covers one subtopic completely enough to be cited on its own.
Is Your Content Structured So an AI Can Lift It?
Yes — structure is the difference between a page the AI reads and a page it quotes. Each section should stand alone: a definition or answer, the supporting facts in the same passage, and tables or lists where a comparison or sequence is involved. If a paragraph only makes sense after reading the three before it, it cannot be lifted.
Google asks for ordinary editorial structure — paragraphs, sections, clear headings — and explicitly rules out the extreme version: no requirement to break content into tiny pieces, and no ideal page length. A comparison table beats three paragraphs of “on the other hand,” because it compresses exactly the structure an AI answer needs. The same shape travels beyond Google — see our ChatGPT citation playbook.
- ☐ Each section opens with the answer and keeps its supporting facts in the same passage.
- ☐ Comparisons are tables; sequences are numbered lists.
Does Your Structured Data Match What Readers See?
Structured data helps machines understand your page, but Google’s rule is plain: it must match the visible text. There is no special schema for AI Overviews or AI Mode, and inventing one — or marking up claims your page does not make — is worse than adding nothing. Use honest Organization, Article, and FAQ markup that mirrors what readers see.
Google’s May 2026 guide is blunt: structured data is not required for generative AI search, and there is no special schema.org markup you need to add. It remains worth using as part of ordinary SEO, since it supports rich-result eligibility. Keep it simple — Organization schema with accurate sameAs links, Article schema with real author and dates, FAQ schema only on questions the page genuinely answers.
- ☐ Every structured-data claim matches visible text on the page.
- ☐ Organization and Article markup use real names, real profiles, and real dates.
Can a Reader Tell Who Wrote This — and Why to Trust It?
An AI answer is a trust decision as much as a retrieval decision: the model links sources it can present as reliable. Show a real byline that leads to a real author page, cite your sources by name, and write from first-hand experience where the topic allows it. Google’s guidance treats clear authorship and sourcing as core evidence of E-E-A-T.
Google’s helpful-content guidance makes the test practical: is it self-evident who authored the content, does the byline lead to background about the author, and does the page show first-hand experience? Originality is part of the same check: Google’s guide says unique, non-commodity content will influence a site’s presence in AI search more than any other suggestion. If your page adds nothing a generative model couldn’t produce from the top ten results, it has nothing an AI answer needs.
- ☐ Every article carries a real byline linking to a real author page.
- ☐ Claims cite named sources with dates, not vague “studies show” phrasing.
Is Your Page Fresh — Honestly Fresh?
Freshness matters, but only real freshness. AI answers favour current pages for time-sensitive questions, so show visible publication and update dates and revise pages when the facts genuinely change. Google explicitly warns against changing dates to make a page look fresh when nothing substantial changed — that is a quality signal in the wrong direction.
Retrieval systems prefer current sources for questions where the answer moves — prices, policies, features, rules. A visible “Updated November 2026” beside statistics that still cite 2023 data is worse than no date at all, because the model may quote the stale figure as current. Set a review rhythm: check evergreen explainers every few months, and fast-moving topics whenever the facts change. When a page’s answer no longer holds, say so plainly.
- ☐ Publication and update dates are visible on the page and accurate.
- ☐ Every statistic has been checked against a current source.
Does Your Content Cover the Subtopics Around the Question?
A single page that answers one question well is a start; a page that also answers the questions around it is citable far more often. Because fan-out searches subtopics, cover the definitions, the comparisons, the common mistakes, and the “what happens next” on the same page. Depth on one page beats the same answer spread thin across five.
This is where most “AEO-optimized” pages quietly fail: they answer the head question in a paragraph and stop, leaving the fan-out queries to competitors. A genuinely useful page anticipates the next three questions — what it costs, what goes wrong, how it compares — in short paragraphs under their own headings. Choose subtopics from your own data: Search Console queries and buyer questions.
- ☐ The page answers the head question plus its three or four most common follow-ups.
- ☐ Subtopics come from real query and conversation data, not guesswork.
Are You Measuring Citations Instead of Guessing?
Citations are a visibility metric, not a traffic metric, so measure them that way. Track AI Share of Voice — the share of AI answers for your target questions that cite your brand — with a weekly sample of 20–50 real buyer questions in a clean browser session. Search Console folds AI impressions into ordinary organic data.
Google counts AI Overview and AI Mode impressions and clicks inside the Search Console Performance report under the Web search type, blended with ordinary organic data, and has added a Generative AI performance report showing impressions by page and date. Neither shows citation share. Our guide to tracking AI Overview citations lays out the spreadsheet method: a weekly question list, run signed-out, logging appearance, citation, and cited pages.
Trend is everything: react to four-week trends, not single data points. A rising presence rate with flat citations is the warning sign.
- ☐ A question list of 20–50 real buyer questions exists and is run weekly.
- ☐ AI Share of Voice is tracked as a trend, with competitors on the same questions.
What Should You Stop Doing?
Plenty — and most of it is sold as AEO. Google says it ignores llms.txt files and offers no way to request inclusion or pay for a citation. Google-Extended controls AI training and grounding in some other Google systems, not AI features in Search. If a tool promises guaranteed citations, it is selling something it cannot deliver.
Google’s mythbusting section is the shortest route to a sane AEO budget: no llms.txt, no chunking, no rewriting in a special “AI style,” no chasing inauthentic mentions — its spam systems apply to generative AI responses too. There is no ideal page length, no secret schema, and no submission form. Every hour spent on an AI-only file is an hour not spent on the one thing Google’s guide says matters most: content people find unique, compelling, and useful.
- ☐ No llms.txt file, AI-only markup, or “chunked” rewrite.
- ☐ No budget goes to guaranteed-citation or bought-mention schemes.
The Checklist at a Glance
- Indexed, snippet-eligible, and included in Search generative AI features.
- Direct answer in the first paragraph under every question heading.
- Headings as real questions, covering the subtopics fan-out will search.
- Self-contained passages, tables for comparisons, lists for sequences.
- Structured data that matches the visible text — nothing invented.
- Real byline, real author page, named sources, first-hand experience.
- Visible, accurate dates; facts re-checked on every update.
- Subtopics from real query data, answered on the same page.
- Weekly Share of Voice tracking; monthly Generative AI report review.
- No llms.txt, no chunking rewrites, no bought mentions, no guaranteed-citation promises.
Why Bother? What Citations Are Actually Worth
Before investing, it is worth knowing what a citation buys. In a field experiment by Saharsh Agarwal and Ananya Sen, researchers randomly hid AI Overviews for some searchers and not others. Conditional on an Overview appearing, its presence cut outbound organic clicks by 39.8% and raised zero-click searches by 34.5%. Citations inside the Overview generated only about 8% of all clicks, according to the authors’ own summary. Our full breakdown of what AI Overviews do to clicks covers the design and the limits.
Those limits matter: this is a preliminary working paper, not a peer-reviewed study, run on US desktop Chrome users over two weeks in January and February 2026. Treat 39.8% as a causal estimate under those conditions, not a universal rate. The citation is brand exposure at the exact moment of a question — not a replacement for the visit. Plan AEO as a visibility play, judged on Share of Voice.
FAQ
Do I need special schema to appear in AI Overviews?
No. Google states structured data is not required for generative AI search and no special schema.org markup exists to add. A page must be indexed, snippet-eligible, and the site included in Search generative AI features. Ordinary schema still helps rich results — and must match the visible text.
Will an AI Overview citation replace my lost clicks?
Not in the evidence we have. In the Agarwal and Sen experiment, citations generated only about 8% of all clicks, per the authors’ summary, while the Overview’s presence cut outbound clicks by 39.8% where it appeared. Treat a citation as visibility, not traffic.
Does blocking Google-Extended remove my content from AI Overviews?
No. Google-Extended controls AI training and grounding in some of Google’s other systems, not AI features in Search. To limit what AI Overviews and AI Mode can use, use nosnippet, data-nosnippet, or max-snippet; to remove the page entirely, use noindex.
Should I create a separate page for every fan-out query?
No. Google’s guide warns that creating separate content for every query variation to manipulate rankings or AI responses violates its scaled content abuse policy — and calls it ineffective, since its systems understand relevance without exact matches. Cover subtopics as sections on one strong page.
How long does it take to start getting cited?
There is no fixed timeline, and anyone quoting one is guessing. Technical prerequisites resolve in days once fixed; content and authority work compounds over months. Track Share of Voice weekly and judge the trend over a quarter, not a week.
Bottom Line
Answer engine optimization is not a new discipline with secret rules. It is ordinary SEO, done with unusual discipline: an indexed page, a direct answer under a real question, passages a machine can lift intact, honest markup, a real author, current facts, and measurement that tracks citations rather than wishes. Google has written most of this down.

