To rank in AI Overviews and get mentioned by ChatGPT, do eight things: answer the query in your first 60 words, give every section one quotable passage, add FAQPage and Article schema, build a verifiable entity for your business, publish llms.txt, earn mentions on pages AI already trusts, keep your dates honest, and measure AI referrals so you know what worked.
That is the whole method. The rest of this page walks through each step in enough detail to hand to whoever runs your website: no vendor jargon, no subscription required. I run this exact sequence on my own sites, so every step below is one I have personally implemented, not summarized from someone else's deck.
One ground rule before the steps: AI engines mostly quote pages that already do well in normal search. Google's AI Overviews are assembled largely from high-ranking results, and ChatGPT's browsing leans on conventional search indexes. So none of this replaces the basics: a crawlable site, decent load times, real content, some links. If those are broken, fix them first. The definitional version of this discipline, with the full checklist and its limits, lives on the answer engine optimization hub; this page is the hands-on twin.
The eight steps
Answer the query in the first 60 words
Whatever question a page targets, answer it completely in the opening paragraph: 40 to 60 words, directly under the headline, before any backstory. AI engines pull short passages, and the top of the page is where they look first. A page that spends four paragraphs warming up has nothing liftable where it counts.
The self-test: read only your first paragraph. If it fully answers the question the headline asks, you pass. This page's own first paragraph is the demonstration.
One quotable passage per section
Repeat the same move at every H2: open each section with two or three sentences that answer the sub-question the heading raises, and that make sense with everything around them deleted. A page with eight sections built this way offers an engine eight candidate quotes instead of one.
Write the passage first, then add the supporting detail beneath it. If a section cannot produce a standalone claim, merge it into one that can.
Add FAQPage and Article schema
Schema is structured markup (a block of JSON in your page's code) that tells machines what the page is. Two types do most of the work: Article (who wrote this, for which organization, published and updated when) and FAQPage (these exact questions, with these full written answers). Put the complete answer text in the markup, not a teaser.
Add HowTo markup on step-by-step pages (this page carries one) and BreadcrumbList everywhere. Any developer can install these in an afternoon; templates make it a one-time cost.
Build your entity
An "entity" is the machine-verifiable identity of your business. Engines will not name a business they cannot corroborate, so give them the trail: an about page that plainly states who runs the company and their history; Organization and Person schema; sameAs links pointing at your LinkedIn, YouTube, and other profiles; and a Wikidata item if your business has the third-party coverage to qualify.
Then enforce consistency. Same business name, same one-line description, everywhere it appears. Every mismatch is a reason for the engine to skip you.
Publish llms.txt
llms.txt is a plain-text file at yourdomain.com/llms.txt that tells AI crawlers, in ordinary sentences, what your business is and which pages answer which questions. Think of it as the front desk for machine visitors: instead of inferring your site's structure, they get told.
Keep it short, factual, and current. A description that contradicts the live site teaches crawlers to distrust both. Ours took an hour to write and minutes a month to maintain.
Earn mentions on pages AI already trusts
Engines corroborate before they quote. A business mentioned on industry publications, credible directories, and reference pages gets named; a business that exists only on its own domain rarely does. This is classic PR wearing a new hat: guest contributions, expert commentary, real directory listings, coverage.
Prioritize the pages engines already cite in your category: ask the engines your target questions and note which third-party sources keep appearing. Those are your outreach list.
Keep dateModified honest
Engines prefer fresh sources, and your page's dateModified is how you declare freshness, but they also compare page versions. Bumping the date without changing the content is detectable and costs trust. The honest version wins on both counts: review your answer pages quarterly, update the facts that moved, and let the date reflect real edits.
Set a recurring calendar block for it. Freshness is a habit, not a project.
Measure AI referrals
Build a GA4 segment for sessions arriving from chatgpt.com, perplexity.ai, copilot.microsoft.com, and gemini.google.com. Watch Search Console for pages whose impressions climb while clicks stay flat, a common signature of appearing inside AI Overviews. And once a month, ask the engines your target questions yourself and log who gets cited.
The numbers will undercount (many AI answers send no referrer), so read them as a trendline, not a ledger. Direction is enough to steer.
What doesn't work
Three tactics that feel like AI optimization and reliably fail: stuffing pages with "for LLMs" keyword blocks, faking freshness by bumping dates on unchanged content, and mass-publishing thin AI-generated pages to blanket every question in your category.
The keyword-stuffing failure is structural: answer engines retrieve meaning, not strings, so a paragraph written for a parser instead of a reader scores worse on the exact extractability the engine selects for. Fake freshness fails because engines diff page versions: a moving date on static text marks the whole domain as gaming the signal. And thin mass-generated pages fail twice: they rarely rank well enough to enter the candidate pool, and when engines do sample them, generic content gives nothing worth quoting. Google's spam systems have also been explicit about demoting scaled low-value content, whoever or whatever wrote it.
Notice what all three have in common: they try to signal quality without producing it. The engines are imperfect, but they are calibrated on exactly this distinction. The tooling question is separate from the judgment question. AI in this workflow is genuinely useful for the sweep work, auditing schema across hundreds of pages or flagging stale claims, which is how one experienced marketer now maintains what used to take a team. What it does not supply is the judgment about what your pages should claim and which questions deserve a page. That part stays human, and it is the part the engines end up quoting. The same division of labor holds across the basics; see marketing fundamentals for the wider version of that argument.
How long it takes
Steps 1–5 and 7 are within your control and take days to weeks: a content pass over your key pages, a schema install, an about-page rewrite, one text file. Step 6, earned mentions, takes months, because other people's pages move at other people's speed. Expect the on-site work to show up in AI answers unevenly: informational pages first, commercial mentions later, and nothing on a guaranteed schedule.
That lag is not a reason to wait. The candidates being quoted in your category a year from now are being selected from work that is live today. The broader context (how much of search this becomes and what happens to the click economy underneath it) is mapped on the future of SEO and zero-click search pages.
AI search FAQ
How do you rank in Google AI Overviews?
Google AI Overviews assemble answers mostly from pages that already rank well for the query, so the path is: earn a top-20 organic position, then make your page the easiest one to quote. That second part means opening with a direct 40–60 word answer, giving every section a self-contained passage, carrying complete Article and FAQPage schema, and presenting an unambiguous author and organization the system can verify. There is no separate submission process and no guarantee: you are raising the probability of selection, and the sites that get cited consistently are the ones that made extraction effortless.
How do I get my business to show up in ChatGPT?
ChatGPT mentions businesses it can find and corroborate: through its browsing of live pages that rank for the question, and through the mentions of your business already embedded across the web. To show up, make your site crawlable to AI user agents, publish pages that answer the questions your buyers actually ask in their own words, keep your name and description identical across your site, LinkedIn, directories, and profiles, and earn mentions on industry pages the model already trusts. A clear entity with consistent corroboration gets named; an ambiguous one gets skipped.
How do I rank in AI search results?
Across AI Overviews, ChatGPT, Perplexity, and Copilot, the same eight moves apply: answer the query in the first 60 words; give every H2 one quotable passage; add FAQPage and Article schema; build your entity with an about page, sameAs links, and Wikidata where possible; publish llms.txt; earn mentions on trusted pages; keep dateModified honest; and measure AI referrals in GA4. The engines differ in how they retrieve, but they all reward the same underlying properties: extractable answers, verifiable structure, a clear entity, and real authority.
Does traditional SEO still help with AI search?
Yes — traditional SEO is the entry ticket. AI Overviews draw heavily from pages that already rank, ChatGPT's browsing consults conventional search indexes, and Perplexity cites pages that would sit on page one anyway. Crawlability, site speed, internal linking, backlinks, and keyword-relevant content all still determine whether you are in the candidate pool at all. What changes is the last mile: ranking used to be the finish line, and now it is the qualifying round. The pages that win the citation are ranked pages that are also structured for extraction.