You typed a question about your own trade into Google last week, and Google answered it on the results page: a few paragraphs at the top, three or four links tucked beside them, and no reason to scroll. If none of those links were yours, you already know the problem. To rank in AI Overviews, you need to know which of your customers’ searches trigger one and what Google says decides whether your page gets used. The 90-day sequence at the end is one a plumbing company, a body shop or a law firm can run without buying another platform.
What AI Overviews are, and what “ranking” in one means
An AI Overview is the AI-written summary Google puts above the blue links on some results pages, with a handful of supporting links to the pages it drew from. Google describes them as helping people “get to the gist of a complicated topic or question more quickly” and as “a jumping off point to explore links to learn more” (AI features and your website, Google Search Central). AI Mode is the conversational counterpart, meant for questions “where further exploration, reasoning, or complex comparisons are needed”.
Strictly speaking, an AI Overview has no rank and no position one. Your page is either used as a supporting link or it isn’t, and Google notes that AI Overviews “are only shown when our systems determine that it is additive to classic Search, and as such, often don’t trigger”. That leaves you with two questions to work on: is your page eligible, and is it the most liftable answer on the web to that particular question?
Eligibility, at least, has a written bar. To be shown as a supporting link, Google says a page “must be indexed and eligible to be shown in Google Search with a snippet, fulfilling the Search technical requirements. There are no additional technical requirements.”
Which searches trigger an AI Overview, and which of yours will
Google doesn’t publish a trigger list. Its AI features page says AI Overviews “were designed to show up on queries where they can add additional benefits beyond what people might already get on Search”. Its advice to site owners adds that people using these surfaces are “asking longer and more specific questions — as well as follow-up questions to dig even deeper” (Top ways to ensure your content performs well in Google’s AI experiences on Search).
In practice, that means you stop treating your keyword list as one list and sort it into three buckets:
| What the customer types | What usually answers it | What wins it |
|---|---|---|
| “plumber near me”, “body shop open now” | The map pack and local listings | Business Profile, reviews, local search work |
| “how much does a water heater replacement cost”, “paintless dent repair vs respray” | An AI Overview, then the blue links | A clean, extractable answer on a page that already ranks |
| “is Acme Plumbing licensed”, “Acme Plumbing reviews” | Brand results, directories, review sites | Third-party pages you don’t control |
Answer engine optimization works on the middle row. The top row is still won the old way, and if you haven’t done that work yet, start with local SEO for a small business instead. An AI Overview strategy built on top of an unclaimed Business Profile wastes a quarter.
How Google picks the pages it cites
Both AI Overviews and AI Mode “may use a ‘query fan-out’ technique — issuing multiple related searches across subtopics and data sources — to develop a response” (AI features and your website). While the answer is being written, Google says its models “identify more supporting web pages”, which is why these boxes can show “a wider and more diverse set of helpful links” than a classic results page.
Read that as a two-stage job. First retrieval: one of those fanned-out sub-searches has to surface your page. Then generation: the model has to find a passage that answers the sub-question cleanly enough to summarize and link. A page that ranks but buries the answer in paragraph nine loses at the second stage; a perfectly worded answer on a page nothing can find loses at the first.
That is also why Google offers no special lever for these features: “The best practices for SEO remain relevant for AI features in Google Search … There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” Treat AEO as SEO with a stricter editing standard.
Step 1: earn the classic rankings first
A good share of what gets cited still comes from the classic rankings. In Ahrefs’ March 2026 study of AI Overview citations, 38% of the cited pages also ranked in the top 10 for the same query. Ahrefs’ likely explanation for the rest is query fan-out, where Google splits the question into related sub-queries and cites the pages that turn up most often in those results. The practical rule is still unglamorous: if you are nowhere near the first page for “water heater replacement cost” in your city, or for the smaller questions around it, the fan-out has little reason to pick you.
So the first AEO task is ordinary search engine optimization: a crawlable site, pages that target the questions you want to be known for, and enough authority that Google trusts yours over the other shops in your metro. If you’re weighing whether this replaces the SEO you already pay for, we laid the two side by side in generative engine optimization vs traditional SEO.
Step 2: write answers that can be lifted straight off the page
You control this part completely, and in our experience it’s where many small-business sites fall short. A page written for a machine to skim reads differently from a page written to sell.
Lead every section with the answer
Put the direct answer in the first two or three sentences under the heading, in plain language, with the specifics in it. “A standard 50-gallon gas water heater replacement in the Denver metro usually runs between X and Y, depending on venting and whether the unit has to be moved” is liftable. “We pride ourselves on transparent pricing” is not.
One question per heading
Headings that are the question your customer typed give the retrieval stage something to match. “How long does a bumper respray take?” beats “Our process”.
Use formats that survive extraction
Short lists, small comparison tables, definition sentences and numbered steps all survive being pulled out of context. Long, story-shaped paragraphs do not. Keep the story, and put it after the answer.
Put the facts in text
Google’s fundamentals list includes “making sure that important content is available in textual form”. Price ranges, service areas, hours, license details and process steps belong in HTML, not baked into a hero image or a PDF.
Step 3: clear the technical bar, and check you haven’t opted out
Google’s list of what still matters for AI features is short and dull: crawling allowed in robots.txt “and by any CDN or hosting infrastructure”, content findable through internal links, a good page experience, important content in textual form, structured data that matches the visible text, and Business Profile information kept up to date.
Two self-inflicted problems are worth checking today. The first is preview controls. Google’s guidance on AI search is explicit that nosnippet, data-nosnippet, max-snippet and noindex are how you manage what appears, and that “more restrictive permissions will limit how your content is featured in our AI experiences”. If a theme or a previous developer set max-snippet:0 site-wide, you are out of the running and nothing on the site tells you so.
The second is newer. Search Console now has a Search generative AI control under Settings, rolled out to all websites worldwide on August 31, 2026. The default is to include your site; the exclude option prevents your links and content from appearing in AI Overviews, AI Mode and generative AI features in Discover, and Google states that you then “won’t receive any traffic or impressions from these features”. Check which one your property is set to before you spend a quarter on content.
The same audit applies to the engines that aren’t Google. OpenAI documents OAI-SearchBot as the crawler behind ChatGPT’s search features and warns that “sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links” (Overview of OpenAI Crawlers). Perplexity says PerplexityBot is what surfaces and links websites in its search results, and recommends allowing it if you want to appear there (Perplexity Crawlers). Both publish IP ranges so a firewall doesn’t block them by accident. Making sure these crawlers can read your site is part of our AEO services, because a blocked crawler makes every other effort pointless.
Step 4: structured data that matches what’s on the page
Google says what you can skip: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.” Anyone selling you an llms.txt package as the key to AI Overviews is selling you a file Google has said it does not need.
Structured data still earns its place, because it states your facts unambiguously and makes you eligible for rich results. LocalBusiness markup requires only a name and an address, with telephone, opening hours, geo coordinates, price range and URL among the recommended properties. Google recommends ratings only for sites that review other local businesses. Add Organization for the business and Article with a real author for your guides.
The one rule that isn’t optional is that markup has to match what a visitor sees. Google’s advice for AI search puts it as “making sure that all the content in your markup is also visible on your web page”. Marked-up hours that contradict your Business Profile can do more harm than no markup at all.
Step 5: build a cluster of pages around each service
Query fan-out means a single “services” page is being asked several questions it was never written to answer. The fix is a small hub-and-spoke set: one thorough page per service, plus a short article for each real question customers ask about it, all linked to each other and back to the service page.
A collision repair shop might run one estimate page plus separate pieces on insurance claims versus paying out of pocket and whether OEM parts matter. Together they give Google several pages to draw on when someone asks about collision repair, and a reason to treat you as a source on the subject.
Step 6: E-E-A-T signals a machine can check
Google’s content guidance centers on experience, expertise, authoritativeness and trustworthiness, and is blunt about the ranking among them: “Of these aspects, trust is most important” (Creating helpful, reliable, people-first content). For a small business, that comes down to the signals a machine can verify:
- A named author with a real bio, role and credentials. A post signed “admin” or “the team” gives a machine nothing to check.
- License, certification and insurance details stated on the site and matching the state register.
- Visible published and updated dates, on content that is actually updated.
- An address and phone number that match your Business Profile and directory listings, character for character.
- Claims backed by a link to the source, and photos of your own work rather than stock.
Step 7: off-page mentions, lists and reviews
The fan-out also reads the directories, review platforms, trade association pages, local news and forum threads that mention you. In our experience, that off-page half often decides whether an AI answer names a small business at all.
Claim and correct every listing carrying your name and address. Ask satisfied customers for reviews on the platforms your category actually uses. Get onto genuine local and trade lists. This is familiar listing and review work, and these pages now feed the answer itself, on top of whatever value they had as backlinks.
How to measure whether any of this is working
Since August 31, 2026, Search Console has offered a generative AI performance report to all websites worldwide. It shows impressions for AI Overviews and AI Mode, broken down by page, country, date and device, drawn from the Web search type in the main performance report. If two of your pages appear in the same AI feature, they count as a single impression in the chart total, which is worth knowing before you panic about a flat line.
The report has two limits. It is impression-led, so it tells you that you were shown; as of September 2026 it doesn’t say what was asked or how many people clicked. It also covers Google only. ChatGPT, Perplexity and Gemini report nothing, so you test those by hand, using the method in how to track brand mentions in AI search.
Watch the right outcome, too. Google’s advice on succeeding in AI search reports that “when people click to a website from search results pages with AI Overviews, these clicks are higher quality”, and says you can miss what those visits are worth “if you focus too much on clicks instead of the overall value of your visits”. Track calls, form fills and booked jobs.
A 90-day sequence your team can run
You can run this without a platform subscription, as long as somebody owns it for an hour or two a week.
| Window | What you do | What you have at the end |
|---|---|---|
| Days 1–30 | Audit eligibility: indexing, robots.txt, firewall rules, snippet controls, the Search generative AI control, Business Profile accuracy. Write down the 20 questions customers ask. | A site that can be cited, and a question list ranked by closeness to a sale. |
| Days 31–60 | Rewrite your top service pages answer-first. Publish four or five question pages. Fix Organization and LocalBusiness markup. Run a baseline prompt test across ChatGPT, Perplexity, Gemini and Google. | A cluster around one service, and a baseline of how often you are mentioned. |
| Days 61–90 | Fix listing inconsistencies, chase reviews, pursue two or three genuine third-party mentions. Publish the next batch of questions. Re-run the prompt test. | A month-over-month comparison, and a read on which questions you now own. |
Resist doing this for every service at once. One service done properly teaches you more than five done thinly, and gives your team a template to repeat.
Common questions about how to rank in AI Overviews
Do I need to add llms.txt or special AI markup?
No. Google states that you don’t need new machine-readable files, AI text files or markup, and that there is no special schema.org type for these features. Spend the time on the answer-first rewrite instead.
Will an AI Overview kill my traffic?
Some informational queries will send fewer clicks, because the answer is already on the results page. Google’s position is that the clicks that do arrive are higher quality. Judge it on calls, quotes and booked work over a quarter rather than sessions over a week.
Can I stop Google using my content in AI Overviews?
Yes, through the Search generative AI control in Search Console settings, or by restricting previews with nosnippet, data-nosnippet, max-snippet or noindex. Understand the trade: Google says excluding your site means no traffic and no impressions from those features at all.
How long does this take?
Technical eligibility fixes land as soon as Google recrawls; content and off-page work compound over months. Google guarantees neither indexing nor serving.
Is this just SEO with a new name?
Mostly. The eligibility bar, the crawling, the authority and the content quality are the same work; what is new is the editing standard and the measurement. The full comparison is in generative engine optimization vs traditional SEO.
Does any of this help with ChatGPT and Perplexity?
In our experience, largely yes. They crawl with their own bots, and the clear answers and third-party mentions that help with Google tend to help there too. The differences are which bots you must allow and how you measure, since neither gives you a Search Console.
What to do next
Make sure Google is allowed to use your pages, rank for the question and the smaller questions around it, and answer it in the first three sentences under a heading that matches how a customer asks it. The rest of this playbook applies that to structured data, clusters, trust and off-page mentions.
If you’d rather not run it yourself, that audit-and-rewrite sequence is what our answer engine optimization service covers, and it works best alongside the SEO that feeds it. Tell us which questions you want to own and we’ll come back with a plan that starts with an audit of who the AI answers are naming today.
