Should I optimise for ChatGPT or Gemini?

The gap in our own data, with the denominator

The two engines reached different conclusions about the same business in the large majority of cases. Every figure here is counted over the 137 businesses in our 300-business scan that have complete records from both engines, scanned on 30 July 2026.

Recomputed from data.csv. Shares are rounded and total 100%.

On the scores rather than the naming, Gemini was higher for 105 of the 137 and ChatGPT was higher for 7, which is 77% against 5%, with the remaining 25 tied. Across all 137 the mean absolute gap between the two per-engine scores was 25.3 points out of 100. In the 105 cases where Gemini was actually higher, the mean gap was 31.2 points. A quarter of the scale, between two engines, about the same business, on the same day.

Why Gemini probably scores higher, and why that is not favouritism

The most likely explanation is mechanical: the two engines were asked through different interfaces with different retrieval behaviour. We are offering this as an explanation, not proving it.

We asked ChatGPT through the OpenAI Responses API with the web search tool enabled, and Gemini through the Generative Language API with Google Search grounding. A grounded Gemini answer is assembled from Google's search results, and Google's index of small Australian businesses is deep. If one engine reaches for a search index by default more often than the other does, it will name more small local businesses, and that is exactly the shape of our result.

What this does not mean is that Gemini likes you. It means the question was routed differently. If ChatGPT's search behaviour changes, our numbers will change with it, which is the ordinary condition of measuring something that is being rebuilt underneath you.

What being named by only one engine means for you

It means you have a specific, findable gap rather than a general problem, and the two gaps have different fixes.

Named by Gemini and not by ChatGPT, the position of 26 of our 137, usually means your facts are findable through search but the corroboration is thin outside it. ChatGPT is less likely to volunteer you unprompted. Check first that your robots.txt does not block OpenAI's crawlers, then look at whether anything other than your own website describes what you do.

Named by ChatGPT and not by Gemini, which was 10 of the 137, more often points at the search layer: listings that disagree with each other, a Google Business Profile that is thin or wrong, or pages that are not being indexed. The practical checklist for that side is in how to show up in Google AI answers.

Named by neither, the largest group at 56 of the 137, is not a per-engine problem at all. Start with why doesn't ChatGPT mention my business.

What optimising for both actually looks like in practice

Almost entirely the same work, measured twice. There is no separate ChatGPT strategy and Gemini strategy worth paying for, and anyone selling you two is selling you one twice.

• Let both read you. One robots.txt governs all of them. See should I block AI crawlers.
• State your facts once and consistently. On your site, in structured data, and everywhere else you are listed.
• Earn mentions on pages you do not own. This feeds both engines, because both are reading the same web.
• Measure both, on the same questions, on a schedule. This is the only step that is genuinely per-engine, and it is measurement rather than optimisation.

The manual version of that last step takes about five minutes per engine and is written out in how to check what ChatGPT says about my business.

Where Perplexity and Copilot fit, and why we do not score them yet

They matter, and we do not score them, and we would rather say that plainly than imply a coverage we do not have.

Our deployed scanner queries two engines: ChatGPT and Gemini. Adapters for Claude, Perplexity and Grok exist in the code but have no API key configured, so a report marks them as not configured rather than quietly scoring them as absent. The scanner publishes a live configuration endpoint so that which engines are switched on is a fact anyone can re-check, and it is linked from The Receipts Engine.

In the meantime, checking Perplexity by hand is worth your time. It cites its sources prominently, which makes it the fastest way to see which pages the answer in your category is actually being written from. That list is useful whichever engine you care about.

Where to start

The answer to the question in the title is both, and the only way to know which one is failing you is to ask them the same questions on the same day. Our free scan does exactly that, and you keep the transcripts.