How to Rank in ChatGPT
A practical playbook for getting your brand cited and recommended inside ChatGPT and other generative AI assistants.
Cogniva is a Melbourne-based search visibility and AI systems consultancy. Because we build AI agents and automation systems as well as doing search work, our understanding of how models like ChatGPT retrieve and select sources comes from working with the underlying mechanics — retrieval, ranking, extraction, synthesis — rather than from marketing summaries of them. We help Victorian businesses understand where they actually stand inside AI answers, and what realistically changes it.
There's a quiet assumption baked into the phrase "how to rank in ChatGPT," and it's wrong in a way that matters. The assumption is that ChatGPT is a search engine — that it reads the live web, weighs your page against competitors, and returns the best ones, just with a chat interface bolted on. If that were true, you'd optimise for it roughly the way you optimise for Google.
But that's not what's happening under the hood, and the gap between the assumption and the reality is exactly where businesses waste effort. So instead of giving you tactics, this piece does something more useful first: it explains what ChatGPT is actually doing when it decides whose content goes into an answer. Get the mechanism right and the tactics become obvious. (For the strategy and budgeting layer above this, see our GEO Melbourne pillar.)
1. How ChatGPT actually sources an answer
Start with the single most misunderstood fact: most of the time, ChatGPT isn't looking anything up. It has two fundamentally different ways of answering, and only one of them touches the live web.
In its default mode, ChatGPT answers from memory — the statistical patterns absorbed during training. It doesn't retrieve, look up, or access any external source; it generates responses entirely from parametric knowledge (ZipTie, 2026). No URLs, no live pages, no citations. When it produces something that looks like a citation in this mode, it's reconstructing the shape of a reference from training patterns, not checking that anything exists. This is why ChatGPT sometimes confidently cites papers that were never written.
In its browsing mode, it actually queries the web before answering. And here's the technical fact that reframes everything: ChatGPT's live browsing tool uses Microsoft Bing, not Google, as its search backbone (Webintellitech, 2026) — a consequence of OpenAI's partnership with Microsoft. When browsing fires, it generally doesn't send your customer's messy conversational question straight to search. It appears to first reformulate the question into a cleaner, more specific search phrase — stripping the conversational language down to the core informational intent — then retrieve pages for that reformulated query. OpenAI's own documentation notes that ChatGPT search typically rewrites a query into one or more targeted queries before sending them to providers (OpenAI Help Center).
Which mode runs isn't random, though the exact trigger logic is OpenAI's and not fully documented. In observed behaviour, the model appears to decide whether live information is required based on how stable the likely answer is. A well-established question ("what do red blood cells do?") tends to get answered from memory because the answer barely varies across the training corpus. A volatile or current one ("what happened in the news today?") tends to trigger a search. The practical takeaway isn't a precise ratio — it's that a meaningful share of questions never trigger retrieval at all, so for those queries there's no live page selection happening for you to influence.
So the honest one-line model of ChatGPT's retrieval is this: it's often a closed book working from memory, and when it does open the web, it's reading Bing's results — not Google's, and not the whole internet. That distinction alone invalidates a lot of "ChatGPT SEO" advice written as though Google rankings were the lever.
There's one more step that matters, because it's where most pages die. Being retrieved is not being cited. The figures here come from third-party SEO analyses rather than OpenAI itself, so treat them as directional rather than precise — but the direction is striking and consistent across studies: one analysis of several hundred thousand pages reported ChatGPT citing only a small fraction, on the order of 15%, of the pages it retrieves, discarding the rest after evaluation (AirOps analysis, via Erlin, 2026). Whatever the exact number, the principle holds: retrieval gets you into the room, and something else gets you named.
If you take one mental model from this whole article, make it this: ChatGPT doesn't rank pages — it reconstructs a consensus answer from a small subset of retrievable, corroborated sources. Everything below is a consequence of that one sentence.
2. What gets cited most often
If only a sliver of retrieved pages survive into the answer, the obvious question is: which ones? The patterns here are clearer than you'd expect, and several of them are counterintuitive.
Mid-authority pages punch above their weight.
The intuitive assumption is that the biggest domains win. The data suggests they don't, not proportionally. High-authority sites appear to get retrieved more frequently, but observational analyses indicate they aren't selected at proportionally higher rates — once retrieved, mid-authority pages show citation rates broadly comparable to higher-authority domains (AirOps analysis, via Erlin, 2026). Retrieval favours the giants; selection looks more democratic. That's encouraging for a smaller business — you don't need to be Wikipedia to be cited, you need to be retrievable and then cleanly usable.
3. What actually moves your inclusion probability
Now the practical part — but note that every lever below follows directly from the mechanism, rather than from generic best-practice. These are the things that change whether ChatGPT names you.
Get retrievable on Bing, not just Google.
Because browsing runs on Bing's index, a page Bing hasn't indexed effectively can't be cited in browsing mode. Submitting your sitemap to Bing Webmaster Tools is among the fastest actions you can take to open citation opportunities, and it's worth confirming Bingbot isn't blocked on important pages (Webintellitech, 2026). Most marketers have spent a decade ignoring Bing entirely. It's now a direct ChatGPT lever, and almost nobody is contesting it.
Front-load the answer.
Given that extraction pulls overwhelmingly from the top, put the direct answer to a likely question in the first third of the relevant section. Lead with the conclusion, then support it — the inverse of how most marketing copy is structured. Question-style headings help the model match your content to the reformulated query it's actually searching.
Build corroboration off your own site — and earn into the cluster.
These are two faces of one lever. The model checks whether the wider web agrees you're real, so presence on the platforms it trusts matters: complete, active review profiles, accurate listings, and your name appearing consistently alongside your category across the sources that cover it. That consistent cross-web mention is also exactly what earns you into a citation cluster. A business with no third-party footprint is asking ChatGPT to vouch for it on its own word, which the model is reluctant to do. You can't fake this, which is precisely why it's defensible once you have it — it's the slowest lever here and the hardest for a competitor to copy.
Keep it current.
ChatGPT's browsing layer shares the broader AI-search recency bias — by one measure, content updated within 30 days receives roughly 3.2 times more citations than stale content (ZipTie, 2026). A page that earned citations six months ago and was never touched again quietly stops earning them.
4. Where businesses get misled: the "ranking" myth
Here's the trap, and it catches sophisticated businesses as easily as naive ones: assuming that a strong Google ranking buys ChatGPT visibility. It's the most natural assumption in the world, and it's mostly false.
The correlation exists but the available analyses suggest it's weak. Top Google rankings appear to lift ChatGPT citation odds somewhat — but the same studies indicate only a small minority of ChatGPT-cited URLs also rank in Google's top 10, and that a substantial share of brands with strong Google rankings have little or no ChatGPT visibility (EMGI analysis, via Erlin, 2026). The precise percentages vary by study and should be read as directional, but the conclusion is robust across them: you can dominate Google for your category and still be close to invisible inside ChatGPT's answers. Strong SEO is a contributing factor, not a guarantee — partly because retrieval runs on Bing, partly because selection applies its own criteria afterward.
This is why "ranking in ChatGPT" is a slightly misleading phrase, and worth retiring in your own head. There is no ranked list. There is no position one. ChatGPT doesn't present a ranked list of pages; it generates a synthesised answer by pulling candidate pages, evaluating them, and selecting which to quote or paraphrase (Erlin, 2026). You're not climbing a ladder to the top slot. You're trying to be one of the three-to-six sources the model decides to use — and whether it uses you depends on retrievability, corroboration, structure, and recency, not on a rank you can watch tick upward.
The practical consequence is uncomfortable but clarifying: you can't manage ChatGPT visibility with a rank tracker, and your SEO agency's ranking report tells you little about it. The only honest measurement is to ask ChatGPT the questions your customers ask and see whether you appear, how you're described, and who's named instead. If the answer is "a competitor, and not us," that's the real scoreboard — and most businesses have never looked at it.
Where this model breaks down
In the interest of not overstating the case — because the whole point of this piece is to be more accurate than the hype, not just differently confident — here's where the "ChatGPT is nothing like Google" framing has limits.
Sometimes it does resemble ranking.
When ChatGPT is in browsing mode, the retrieval step is a fairly conventional search against Bing's index, and Bing's index is built on broadly familiar signals — authority, relevance, freshness, technical health. For the retrieval half of the process, classic SEO thinking applies more than the rest of this article might imply. The divergence is concentrated in the selection step that happens after retrieval, not in retrieval itself.
Sometimes SEO correlates strongly.
For some query types — particularly stable, factual, well-covered topics — the pages that rank well and the pages that get cited overlap heavily, because the same authoritative sources dominate both. The weak overall correlation is an average across many query types; in specific categories it can be much tighter.
Sometimes Bing and Google agree anyway.
The two indexes aren't independent universes. For many queries they surface substantially similar pages, so "ChatGPT uses Bing, not Google" matters most at the margins — for pages where the two engines diverge — rather than being a universal gulf.
And the data underneath all this is young.
Most quantitative claims about ChatGPT citation behaviour come from third-party SEO tools analysing samples, not from OpenAI. The figures move between studies and over time as OpenAI changes the product. The directional findings have been consistent enough to act on; the precise numbers should be held loosely.
None of this rescues the "just rank on Google" assumption — that's still wrong. But the honest version is "ChatGPT partly overlaps with search and partly doesn't," not "ChatGPT has nothing to do with search." Treat the overlap as your foundation and the divergence as the part that needs separate attention.
For the strategy sitting above all this — whether ChatGPT visibility is even worth your spend right now, and how it fits your wider search budget — start with whether GEO is worth it for your business.
Good to know
Frequently asked questions
Straight answers before you book anything — and if yours isn’t here, ask us in the consultation.
Free consultation
Want to know if ChatGPT actually mentions your business?
We’ll test the prompts your customers use and show you exactly how the AI engines talk about you — and your competitors.
Book a consultation