Ask ChatGPT for the best option in your category — best CRM for a small agency, best clinic in Brisbane — and it answers with a shortlist. Three to five named brands, a sentence of reasoning for each, often with cited sources underneath.
Your brand is either on that shortlist or, for that buyer, it doesn't exist. There is no page two of a ChatGPT answer.
If you're reading this, I'm guessing you've already run the prompt and didn't love what came back. The good news: which brands ChatGPT recommends is not random, and it is not fixed. What follows is how the machine actually decides — from OpenAI's own documentation, not folklore — and six steps that shift your odds.
Why should I care?
Fair question — here's the number. In August 2026 OpenAI reported more than one billion weekly active users and more than two million businesses on its products. Even a sliver of those sessions being buying questions makes a channel that answers with shortlists, not links.
And an AI answer fails harder than a results page. Rank sixth in Google and you still get clicks; get left out of a five-name ChatGPT answer and you get silence. That winner-takes-most shape is why generative engine optimization (GEO — optimising to appear in AI answers) exists. SEO spent two decades reverse-engineering one black box; the good news is you get to do it all again.
One scope note before I lump you with six steps: I'm not going to talk about Google's AI Overviews here — that's its own playbook: how to appear in Google AI Overviews. This one is ChatGPT, end to end.
The one principle behind everything
ChatGPT doesn't rank pages. It picks sources it already trusts — and you get recommended by becoming one of those sources, or by being present inside them.
Sit with that — it inverts the Google reflex. Google's question is "which page best answers this query?" ChatGPT's is closer to "what do the sources I rely on say about this topic?" — then it synthesises their answer. Your website is one voice in that room; the review platform that lists you, the Reddit thread that rates you, the listicle that ranks you third often weigh more.
So "ranking in ChatGPT" is two jobs: make your own site worth retrieving and quoting, and be present — accurately, favourably — in the third-party sources it leans on. Every step below serves one of the two.
How ChatGPT actually decides what to recommend
Almost everything OpenAI has said publicly fits on one help-centre page — worth reading closely, because most advice skips it.
ChatGPT has two modes. It can answer from training data (what the model absorbed before its cutoff), or it can search the live web. OpenAI's documentation says only that "ChatGPT will automatically search the web if your question might benefit from information on the web" — that is the entire published trigger; no deterministic rule exists to engineer around. Plan for both modes: training answers reward long-standing, widely repeated reputation; live answers reward crawlability and freshness.
Your prompt isn't the query. When it does search, ChatGPT "typically rewrites your query into one or more targeted queries that it sends those providers". One buyer question fans out into several searches you never see — so you can't optimise for one keyword the way you would a Google results page.
The providers mostly aren't OpenAI. The same page names Bing (Microsoft) and Shopify as third-party search providers. If Bing can't find you, a large part of ChatGPT's live retrieval can't either.
And there's an official ranking statement — vague on factors, precise on entry requirements:
"Ranking in ChatGPT Search is based on a number of factors designed to help users find reliable, relevant information. There is no way to guarantee top placement. However, to be included it is important to allow OAI-Searchbot to crawl your site, and ensure your site host and/or content delivery network allows traffic from our published IP addresses."
— OpenAI Help Center, "ChatGPT Search"
Step 1: Let ChatGPT in
Do the plumbing before anything clever. This is the unglamorous end of how to optimise your website for ChatGPT, and it gates everything else.
- Allow OAI-SearchBot in robots.txt. It's OpenAI's search crawler; if a bot-blocking preset or blanket AI-crawler rule catches it, you've opted out of being cited.
- Let OpenAI's published IPs through. Firewalls and CDN bot protection can challenge crawlers that robots.txt allows. OpenAI publishes the addresses at searchbot.json; hand that list to whoever runs your CDN or WAF (web application firewall).
- Get properly indexed in Bing. Bing is a named provider, yet plenty of teams have never opened Bing Webmaster Tools. Verify the site, submit the sitemap, fix what it flags.
Twenty minutes of checks, and the most common technical reason a brand can't rank in ChatGPT search results is off the table.
Step 2: Be present where ChatGPT already looks
If ChatGPT repeats its trusted sources, the next question is obvious: which sources? There's real data.
Start with Peec AI's analysis of 30 million cited sources across five engines: ChatGPT's five most-cited domains, in order, were Wikipedia, Reddit, Forbes, Techradar and LinkedIn. And Profound's longer study — 680 million citations, August 2024 to June 2025 — shows how top-heavy that trust gets: Wikipedia alone was 7.8% of all ChatGPT citations and 47.9% of its top-10 citation share.
Hold that window in mind, though — the mix moved sharply after it closed. Semrush tracked 230,000+ prompts across 13 weekly snapshots and watched ChatGPT's Reddit citation rate collapse from close to 60% of prompt responses in early August 2025 to around 10% by mid-September, while Wikipedia fell from roughly 55% of responses to under 20%.
The trusted-source game is real — a short list of domains carries a huge share of citations. But the list moves, so don't bet everything on one domain; build presence across the review platforms, comparison sites, publications and communities that cover your category. That list matters more than any global top five — the published research is broken down engine by engine in which sources AI engines actually cite.
Practically, how to get mentioned by ChatGPT mostly reduces to how to get mentioned by its sources: pitch the listicles and comparison round-ups your buyers already trust, keep review profiles active, and show up honestly where your category gets discussed. If you want ChatGPT to recommend your business, make it easy for its favourite sources to recommend you first.
See what AI says about your brand
Run the free audit and see what AI assistants tell buyers about businesses like yours.
Step 3: Make your brand facts boringly consistent
Language models deal in entities — a brand as a bundle of facts: name, category, location, who it serves, what it costs. The pattern practitioners report is that brands whose facts read identically everywhere get described accurately, while contradictory descriptions produce vague or wrong answers. That's informed inference, not a documented ranking factor — but acting on it costs almost nothing.
- One description, everywhere. Same name, category and one-line positioning on your site, LinkedIn, Google Business Profile, directories and review profiles.
- Schema hygiene. Structured data (machine-readable labels for your pages — Organization, Product and FAQ markup) stops crawlers guessing what you are — ChatGPT SEO in the most literal sense: ordinary technical SEO doing double duty.
- Answer-shaped pages. Pose a buyer question as a heading and answer it in complete, standalone sentences: FAQ pages, honest comparisons, plain-English pricing pages.
Step 4: Publish content machines can cite — and keep it fresh
The data is pointed about format. Peec AI's benchmark of a million-plus citations found that in ChatGPT, 52% of cited listicles were cited more than twice — list-format content earns repeat citations. Round-ups, ranked comparisons and "best X for Y" pages are the workhorses.
Freshness is the other half. Scrunch AI analysed 3.5 million citation events and found citation activity halves in roughly 4.5 weeks across platforms — with ChatGPT churning fastest, at 3.4 weeks. Their caveat is worth repeating: that's aggregate persistence across many prompts, not what happens if you rerun one exact prompt. Either way: a cited page is a lease, not a freehold. Update key pages on a schedule, date them visibly, and treat "we published that last year" as a liability.
Step 5: Make a call on llms.txt
You'll hit this suggestion within a week of researching AI visibility, so here's the honest version of a contested question.
The case for: llms.txt, proposed by Jeremy Howard of Answer.AI in September 2024, is a standard file giving AI agents a concise, curated summary of your site — a reasonable idea, trivially cheap to implement.
The case against: Ahrefs studied 137,210 domains with llms.txt files and found 97% of those files received zero traffic in May 2026 — and of the files that were fetched, 96% of requests came from bots, with only 19.5% of fetches from named AI tools. Google's John Mueller has said no AI system currently uses llms.txt, as reported by Search Engine Roundtable. Peec AI's own analysis is literally titled "llms.txt & .md files – Important AI Visibility helper or hoax?"
My call: add it if it takes ten minutes, don't let it displace anything above, and don't stuff it with keywords — nothing published suggests anyone influential is reading it. Cheap to add; unproven for search-and-citation visibility today.
Step 6: Track whether any of it is working
Everything above is a bet until you measure it. Write down the 20–50 prompts your buyers actually ask, run them on a schedule, and record: whether you're mentioned, your share of voice (how often you appear versus competitors), and which sources get cited either way. Those citations are next month's outreach list.
You can run this ChatGPT brand monitoring in a spreadsheet; it works, it's just slow and easy to let slide. An AI visibility platform like Apex automates the prompt tracking and share-of-voice maths across ChatGPT, Gemini, Perplexity and the rest. For a baseline before committing to anything, the free evaluation shows what the engines currently say about your brand.
Why it gets confusing
Read three studies on ChatGPT visibility and you'll meet claims that flatly contradict each other. The biggest is worth defusing, because it's where most bad advice starts.
"ChatGPT is just Bing" versus "it barely matches Bing." Seer Interactive found 87%+ of SearchGPT citations matched Bing's top organic results when the exact question was used as the search query (Google matched 56%) — from a small sample, as Seer themselves flag: 100 queries, just over 500 citations. Then Ahrefs studied 15,000 long-tail queries and found ChatGPT's in-text citations overlapped Google's top 10 just 8% of the time — and Bing's barely more, at 8.1%.
Both are real. The one test to carry with you: ChatGPT's source mix is measured, not fixed. Every stat is a property of somebody's query set, window and phrasing, not of ChatGPT itself. Exact questions steer it toward classic search behaviour; conversational long-tail prompts trigger rewriting and scatter it across the web. Different question set, different ChatGPT.
The Reddit whiplash from Step 2 is the same lesson on the time axis. Which makes the common mistakes easy to name — chasing one leaked "ranking factor" as if a deterministic algorithm had been cracked, stuffing llms.txt because a checklist said to, and treating one month's citation mix as permanent.
Same logic for the question everyone eventually asks: does ChatGPT use Google rankings? No direct evidence says so — Bing is the provider OpenAI names, and Ahrefs measured ChatGPT's in-text citations overlapping Google's top 10 just 8% of the time (12% for AI-cited URLs across assistants generally). What correlation exists is because good sites do well everywhere, not because Google position is an input.
What I'd do first
No recap — just the order I'd run these bets.
- Confirm OAI-SearchBot access and Bing indexation today. Twenty minutes, and every other step depends on it.
- Claim and maintain your Wikipedia presence, where you're genuinely notable. It held 47.9% of ChatGPT's top-10 citation share over Profound's August 2024–June 2025 window — diminished since, still not optional.
- Get into the listicles and review sites your category trusts. List formats earn repeat citations; these are the rooms where ChatGPT forms opinions.
- Make your entity facts identical everywhere. Cheap, permanent, and it compounds with every new mention.
- Refresh your key pages on a rolling schedule. A 3.4-week citation half-life means stale content quietly falls out of answers.
- Monitor monthly, not once. The source mix moves; your strategy should notice.
FAQ: ranking in ChatGPT
Can you do SEO for ChatGPT?
Yes, but it works differently from Google SEO. OpenAI states there is no way to guarantee top placement, so the work is making your site crawlable by OAI-SearchBot, getting indexed in Bing, and earning mentions on the sources ChatGPT already cites. Classic SEO hygiene still helps — it just isn't the whole game.
How does ChatGPT choose which brands to recommend?
It combines training data with live web search. When it searches, it rewrites your prompt into targeted queries, sends them to third-party providers including Bing and Shopify, and ranks results on factors OpenAI describes only as helping users find reliable, relevant information. In practice, brands that appear consistently across trusted sources get recommended most often.
Does ChatGPT use Google rankings?
Not directly — there is no evidence of a Google integration, and Bing is the search provider OpenAI names. Published studies put the overlap between ChatGPT citations and Google's top 10 at only around 8–12%, so ranking well in Google neither guarantees nor rules out appearing in ChatGPT.