ChatGPT’s web search can only use a business’s website if OpenAI’s search crawler is allowed to read it. Gemini and Google’s AI answers draw on Google Search and Google Maps, where a Business Profile counts. The two assistants reach for different evidence, so the same question can name different businesses in each. Getting into one does not get you into the other.
Maybe you asked ChatGPT for the best mechanic near Brunswick. It named three workshops, and yours was not one of them. If you run a small business and have never opened your website’s settings, this is for you.
When ChatGPT searches the web for an answer, it can only use pages its search crawler is allowed to read. A crawler is a program that visits websites and reads them automatically. OpenAI runs a separate crawler for each job and publishes the list.
The one that matters here is called OAI-SearchBot. OpenAI says it is ‘used to surface websites in search results in ChatGPT’s search features’. Sites opted out of it ‘will not be shown in ChatGPT search answers’, the same page says. They can still appear as a plain navigational link.
The opt-out lives in a file called robots.txt. It is a short text file on your website that tells crawlers which parts they may read. A line added years ago by whoever built the site can shut ChatGPT out without anyone deciding to.
Search and training are separate permissions. GPTBot is the crawler that gathers pages for training OpenAI’s models. OpenAI says each setting is independent, so a business can refuse training and still be findable in ChatGPT search. It can also refuse both by accident.
Paid ads are a separate track again. OpenAI runs another crawler that checks pages submitted as ads on ChatGPT. Nothing in OpenAI’s documentation for site owners offers a way to pay to be named inside an answer.
Google rankings do not get you past this gate, because it sits on your own site. You could rank first on Google for your suburb and still be absent from ChatGPT’s answers.
Getting through the gate is only the first step. Microsoft’s search team, writing about how its Bing index is changing to support AI answers, put the shift in two questions. ‘Traditional search asks: which pages should a user visit? Grounding asks: what information can an AI system responsibly use to construct a response?’ Grounding means building an answer from sources the system has found, rather than from memory.
Microsoft says the useful unit is now ‘discrete, supportable facts with clear provenance’. It also says that ‘abstention is a valid outcome when support is missing, stale, or conflicting’. When the evidence about a business is thin or contradictory, the right move is to say nothing about it. Being left out is not always a ranking failure.
That is Microsoft describing its own index, not a description of ChatGPT’s insides. Our guide to checking whether ChatGPT can see your business covers the link between ChatGPT and Bing. It also sets out a fuller test.
OpenAI publishes who may read your site. It does not publish how ChatGPT weighs one business against another, and Google does not publish that for its assistant either. Anyone who tells you the exact formula is guessing.
Gemini is Google’s AI assistant. Google’s help pages say the Gemini app uses public information from Google Search and Google Maps. For places, that includes addresses, websites, ratings and opening hours. Google’s guide to the AI features in Search says they rely ‘on our core Search ranking systems to retrieve relevant, up-to-date web pages’.
Google’s developer documentation describes the grounding step in order. The model decides whether a Google Search would improve the answer. If so, it writes and runs one or more searches, then builds a response with citations back to its sources. That page covers the tool developers build on, not the Gemini app itself.
Google’s AI Overviews are the AI summaries at the top of some search results. Along with AI Mode, they can break one question into several searches, which Google describes as ‘issuing multiple related searches across subtopics and data sources’. Google says there are ‘no additional requirements’ to appear in either, and neither of its guides describes a paid route in.
One line in Google’s guide separates the two sides. It says that ‘Google Business Profiles can help your products and services to be visible in both AI responses and other Google Search results’. Nothing in OpenAI’s published crawler and publisher documentation gives a Business Profile any role.
A Business Profile is the listing that shows your business on Google Maps and in local results. Google says three things decide those results: relevance, distance and prominence. Our guide to how Melbourne businesses show up in Google’s map pack explains each one. Google ties its AI answers to the same search systems, so those local signals sit on its side. OpenAI publishes nothing like them.
Distance is why a suburb changes the answer. Asking for the best mechanic in Richmond and asking for the best mechanic in Melbourne are different questions. You should not expect the same businesses back from Google’s AI for both.
Four things help on both sides, as far as the published documentation goes. Three trace back to what Google or Microsoft has written, and one to published research. Our guide to showing up when customers search with AI covers the groundwork that applies to every AI tool.
Consistent business details come first. Google says it builds its picture of a business from crawled web content, licensed data from third parties, and facts contributed by users. Microsoft asks for clear provenance and treats conflicting evidence as a reason to stay silent. If your trading hours or address differ from one site to the next, both descriptions suggest that counts against you.
Facts stated plainly are easier to use. Microsoft’s phrase, ‘discrete, supportable facts with clear provenance’, is the clearest statement of this in either company’s writing. A sentence that says what you do, where, and for whom can be lifted and credited. A paragraph of slogans cannot.
Headings in the words people use help as well. Published research into which pages ChatGPT cites found that pages whose headings closely matched the question were cited noticeably more often. A heading like ‘How much does a brake service cost?’ beats one like ‘Our pricing philosophy’.
Evidence of you away from your own website counts. Google says prominence is ‘based on info like how many websites link to your business and how many reviews you have’. OpenAI publishes no equivalent, so on its side this is a sound bet rather than a stated rule.
A gap between the two assistants is a clue about which kind of evidence is missing. Nobody publishes a formula, so treat these patterns as the first place to look rather than a diagnosis.
Missing from both often means the business is not clearly established anywhere. The details may be thin, out of date, or different across sites. Microsoft’s point about abstention applies here. A system that cannot support a claim about you may simply leave you out.
Showing in Google’s answers but not in ChatGPT’s points first at the crawler permission. Check whether robots.txt blocks OAI-SearchBot. OpenAI also asks sites to allow requests from its published addresses. So a firewall or security setting can turn the crawler away even when robots.txt looks fine.
Showing in ChatGPT but not in Google’s answers points at your Business Profile and the local signals around it. Look at your category, your reviews, and whether your details match everywhere. The local work behind a Business Profile is covered in our guide to why a competitor keeps winning near me searches.
A business can have both problems at once. Start with the side that looks weakest, then ask both assistants again once it is fixed.
Check whether search crawlers are allowed to read your site. It is the only item in this article that can take you to zero in one assistant, whatever else you do. It is also a five-minute question to put to whoever looks after your website.
Ask them one thing: does our robots.txt, or anything else on the site, block OAI-SearchBot? If it does and nobody meant it to, the block can be lifted. OpenAI says it can take about 24 hours after a robots.txt change for its systems to adjust. Our guide to whether you have accidentally blocked AI crawlers covers which ones to allow and how to check.
Be clear about what that fix will not do. It will not put you in Google’s AI answers. That is a different job on a different surface, built on your Business Profile, your reviews and the sites that mention you.
Consistent details across the web, and the reviews and links behind prominence, are not five-minute jobs. They build over months, and they are the part that CJ Digital’s AI SEO service is set up to work on.
Before any of that, run one comparison yourself. Ask ChatGPT and Gemini the same question about your trade and your suburb, on the same day, and write down who each one names. Two lists that differ tell you which half of the work you are missing. You can do it this afternoon, and it costs nothing.
OpenAI does not say how it picks one business over another. Start with the one thing it does say: its search cannot use a site that blocks its search crawler. If that is clear, compare how plainly and consistently your competitor’s details appear across the web against your own.
No. Nothing either company publishes for site owners offers a way to buy a place inside the answer itself. Advertising is a different product, and ChatGPT ads go through their own checks, so a paid ad is never the same as being recommended.
Not on its own. A high Google position does nothing if your site tells OpenAI’s search crawler to stay out, so the permission has to be right first.
On Google’s side, yes: Google names it as something that can help you appear in AI responses. OpenAI’s published documentation for site owners does not mention it.
A crawler fix is quick: OpenAI puts the adjustment at roughly a day. Matching your details and earning reviews and links is slower. Neither company publishes a timeline for that part, so plan in months rather than weeks.
