How to Rank in ChatGPT, Gemini, Perplexity and AI Overviews
If you want to know how to rank in ChatGPT, start with an uncomfortable truth: there is no single ranking to win. AI search engines do not publish a list you climb. Instead, each engine composes its own answer and cites a handful of sources, so you earn a citation in each engine separately, not one position across all of them.
That is the short answer. Now the practical one.
How to rank in ChatGPT and the other AI engines: the short answer
Every major engine works the same way under the hood. It runs retrieval-augmented generation: it retrieves candidate sources, then synthesises an answer from them. So the work is not gaming one algorithm. It is being a clear, authoritative, retrievable source that an engine wants to quote.
But the engines retrieve and cite differently. ChatGPT, Google’s AI Overviews and Gemini, Perplexity and Microsoft Copilot each reach for different material and apply different checks. That means your effort should follow your audience (the engines your buyers, stakeholders and the press actually use) rather than spreading evenly across all of them.
This article is the practical companion to our explainer on what generative engine optimisation actually is. Read that for the discipline; read this for where to point it, engine by engine.
Why AI search has no single ranking — and what replaces it
Classic SEO gives you one ranked list a person scans and chooses from. AI search gives you a synthesised paragraph that quietly cites two, three or four sources and may never send the reader to a website at all.
Most engines follow a shared pattern. They fan a query out into sub-questions, retrieve candidate pages for each, rerank those candidates, then compose a constrained answer with citations attached. A document usually has to pass relevance, freshness, structure and authority checks before it earns one of those citations.
Because there is no ranking, you measure different things. Share-of-answer: how often you appear at all. Average position when you are present. And how the model frames you, which is really a sentiment question. Tracking those three over time is the heart of our AI search visibility programme.
How to rank in ChatGPT
ChatGPT Search draws heavily on the Bing index, plus OpenAI’s own crawlers, OAI-SearchBot and GPTBot. In practice, its citations tend to track Bing’s top organic results closely.
Three implications follow.
First, do not ignore Bing. Visibility there feeds directly into what ChatGPT can retrieve and quote.
Second, make sure your crawlers are not blocked. Check robots.txt and confirm OAI-SearchBot and GPTBot can reach the pages you want cited.
Third, answer the question high on the page. A clear, self-contained answer near the top of the page is far easier for a model to lift and attribute than the same point buried deep in the copy.
Freshness matters too. Ahrefs research on what ChatGPT cites found that cited pages tend to be recently updated, so keep your authoritative pages current.
For a B2B or considered-purchase audience that leans on ChatGPT to research vendors, this is where to concentrate: Bing-side visibility, open crawlers, and pages that state your facts plainly and stay maintained.
How to rank in Google’s AI Overviews and Gemini
AI Overviews and AI Mode run on Gemini and use Google’s existing search index. That means the same signals that drive organic ranking (relevance, freshness and E-E-A-T) also drive inclusion in the AI answer.
What makes Google distinct is grounding. Gemini cross-checks each generated statement against indexed pages before it attaches a citation, so verifiable, well-attributed claims win. Google describes this multi-stage grounding in its developer guidance on grounding Gemini with Google Search.
High organic ranking helps, but it does not guarantee inclusion. Clarity, neutrality and structured expertise can lift a mid-ranking page into the answer, a point reinforced by Google Search Central’s own guidance on optimising for AI features.
For broad consumer and high-volume informational audiences, AI Overviews dominate the top of the page. If that is who you serve, this is your priority engine.
How to rank in Perplexity
Perplexity uses hybrid retrieval, keyword matching (BM25) alongside semantic embeddings, and then applies multi-layer machine-learning reranking. That reranking structurally favours authoritative sources: earned media, reference sources and Tier-1 publications.
There is also a citation gap. Perplexity visits far more pages than it cites. So passing every filter — relevance, freshness, structure and authority — matters more here than on-page keyword density. A page that clears all four earns the citation; one that clears three gets visited and dropped.
The practical focus shifts off your own site. Earned media and reference-source accuracy carry disproportionate weight in Perplexity’s answers.
For research-led audiences — analysts, researchers and journalists who favour Perplexity — that is where to invest: the third-party record, not just your owned pages.
Microsoft Copilot and the wider engine set
Copilot draws on Bing too, so most of the work you do for ChatGPT carries over to Copilot at the same time. Treat them as a pair rather than two separate projects.
New engines keep appearing, and the citation landscape shifts as they do. That is precisely why a fixed, tracked prompt set across engines beats one-off checks. It tells you what is changing rather than leaving you to guess.
The principle holds throughout: weight your effort by where your specific audience asks, not by which engine is generating the most headlines.
What actually moves the needle across every engine
Strip away the per-engine detail and the common ground is clear.
- Answer the question early and plainly. A direct, self-contained answer near the top of the page is the single most portable advantage across all engines.
- Structure content well. Descriptive headings, short definitional sentences, lists and tables make a page easy to parse.
- Make claims verifiable. Clear attribution, accurate reference sources and freshness all help a model trust your version.
Then there are the off-page signals models lean on: authority and earned media. This is where reputation management work that builds authority and earned media and strategic communications feed directly into AI visibility.
Mind the technical hygiene too. Do not block AI crawlers, keep your structure clean, and use structured data so retrieval is easy.
Finally, measure. Run a fixed prompt set on a regular cadence and track presence, accuracy and citations against a baseline rather than guessing. Watching what each engine says is also how you catch errors early. See when ChatGPT gets your company wrong.
How Morris McLane runs this in practice
Knowing the mechanics is one thing; operating them across engines, week after week, is another. That is the work our AI search visibility programme does.
We start by building a fixed prompt set — the priority questions your buyers, stakeholders and the press actually ask — and run it through ChatGPT, Gemini, AI Overviews, Perplexity and Copilot on a regular cadence. Each run captures the answers and their citations, so we can track share-of-answer, average position and how each model frames you against a real baseline rather than anecdotes.
From there we work the source layer the engines retrieve from. That means correcting inaccurate reference pages, tightening attribution on the claims models lean on, adding structured data so retrieval is easy, and confirming AI crawlers can reach the pages you want cited.
Then we build the off-page record that reranking rewards — earned media and authoritative third-party sources — and watch the prompt set to catch errors and shifts early, feeding what we learn back into the next cycle.
Where to focus by audience
A simple way to decide where to start:
- Consumer or broad informational audiences → prioritise AI Overviews, which dominate those queries.
- B2B, considered-purchase audiences → focus on ChatGPT and Copilot, which means Bing-side visibility.
- Research, analyst and press audiences → focus on Perplexity and the earned-media record it rewards.
Whichever you choose, the underlying work compounds. Structural improvements — accurate reference sources, well-attributed claims, structured data and a healthy citation network — carry across model retraining cycles rather than resetting each time.
The short version
There is no single ranking in AI search. You earn a citation in each engine separately, and each one retrieves and cites differently. Answer questions early, make your claims verifiable, keep your crawlers open and your sources accurate, then weight your effort towards the engines your audience actually uses, and measure what changes.
That is the work, and it compounds. If you want to baseline where you stand across ChatGPT, Gemini, Perplexity and AI Overviews, that is exactly what our AI search visibility programme is built to do.
Frequently asked questions
How do I rank in ChatGPT?
You earn citation rather than a ranking. ChatGPT Search retrieves candidate pages largely from the Bing index and OpenAI's own crawlers, then cites a few sources in its answer. The practical levers are sound Bing visibility, crawlers that aren't blocked, answers placed high on the page, and content kept current and authoritative.
Is ranking in AI search the same as SEO?
They overlap but reward different things. Classic SEO optimises a page to appear in a list a person scans; AI search optimisation (often called GEO) optimises the wider information environment so an engine retrieves your facts and attributes them to you. Clean, authoritative content helps both, but AI engines add their own retrieval, reranking and grounding steps.
Why does my brand appear in one AI engine but not another?
Because each engine retrieves and cites differently. ChatGPT and Copilot lean on the Bing index; Google's AI Overviews and Gemini use Google's index and multi-stage grounding; Perplexity uses hybrid retrieval and reranking that favours authoritative, earned-media sources. A page that satisfies one engine's signals may not pass another's, so coverage varies by engine.
Which AI engine should I focus on?
Weight effort by where your audience actually asks. Broad consumer and informational queries are dominated by Google's AI Overviews; considered B2B research often runs through ChatGPT and Copilot; analysts, researchers and journalists lean on Perplexity. A tracked prompt set across engines tells you where you already appear and where the gaps are.
How do you measure AI search visibility when there are no rankings?
By observation. You run a fixed set of priority questions through each engine on a regular cadence, capture the answers and their citations, and track three things over time: share-of-answer (how often you appear), average position when you do, and how the model frames you. That gives a comparable baseline instead of anecdotes.
Does optimising for AI search keep working when the models update?
Largely, yes. Models retrain by re-ingesting the web, so structural work (accurate reference sources, well-attributed claims, structured data and a healthy citation network) carries forward into each cycle rather than resetting. Engine-specific tactics shift, but the underlying authority and clarity compound.
