Choose the questions
A set of 25 to 40 questions built from how your buyers really ask: sales calls, enquiry forms and Search Console queries. They cover finding, comparing, checking price and choosing.
Research
Better recommendations start with better evidence.
Find out what each AI platform needs to understand your business, which sources support the work and how to measure whether your business is becoming easier to find and recommend.
What is on this page. A plain-language comparison of ChatGPT, Google AI Overviews, Gemini and Perplexity, backed by official documentation. The testing protocol shows how buyer questions, recommended businesses and cited sources are recorded, so you can judge progress from the actual answers.
Platform differences
Four platforms, four different ways a buyer meets your business
The fundamentals are shared. What changes is where the answer appears, which crawler reads your site and what each platform has said about how it works.
| Platform | Where a buyer sees it | What is documented | What it needs from your business | What we check |
|---|---|---|---|---|
| ChatGPT | ChatGPT search answers, with links to the pages it used | ChatGPT looks things up on the web and links the pages it used. OpenAI publishes a separate crawler for search results, OAI-SearchBot, apart from GPTBot, which is for training. Source | OAI-SearchBot allowed in robots.txt, a page that answers the question in plain words, and business details that agree everywhere they appear. | We ask your buyer questions in ChatGPT with search on and save the answer and the sources shown. |
| Google AI Overviews | A summary at the top of some Google results, with links to supporting pages | AI Overviews is a feature inside Google Search. Google says no special markup or AI-specific files are needed to appear in it, and that standard SEO practice still applies. Source | A page Google has indexed and can show with a snippet, written to help the reader, from a business whose details match its Google Business Profile. | We run your buyer questions in Google Search, record when an overview appears and which pages it links. |
| Gemini | Google's assistant, as its own product | Gemini is not the same as AI Overviews. AI Overviews is a part of Google Search, while Gemini is a separate assistant, so a good result in one does not mean a good result in the other. | The same fundamentals as the others: readable pages, consistent business facts and independent sources that agree with you. | We test Gemini on its own and record what it says, rather than assuming it matches Google Search. |
| Perplexity | Answers with numbered links to their sources | Perplexity shows its sources next to the answer. It documents PerplexityBot for its search results and Perplexity-User as a separate agent that fetches a page when a person asks. Source | PerplexityBot allowed in robots.txt, pages that state facts plainly, and independent sources a reader can check. | We ask your buyer questions in Perplexity and save the answer, the businesses named and the sources linked. |
What is the same on every platform
- A page the platform can fetch and read, so crawler access comes first.
- Facts about your business that agree everywhere they appear.
- A clear answer near the top of the page that can be quoted accurately.
- Independent sources that describe the same business in their own words.
Where llms.txt and structured data fit
We publish both, and we use them on client sites. They are supporting infrastructure, not a guaranteed way to be cited. Google says no special markup or AI-specific files are needed to be included in its AI features, and that standard SEO practice still applies.
Sources
- Google Search Central: AI features and your website
- OpenAI documentation: OpenAI crawlers (OAI-SearchBot, GPTBot, ChatGPT-User)
- Perplexity documentation: Perplexity crawlers (PerplexityBot, Perplexity-User)
These are the vendors' own pages. Nothing here claims how any platform ranks businesses internally, and no single source, including Bing, is treated as the one place an assistant looks.
Testing protocol
How we test what the platforms say about a business
The same protocol sits behind the AI visibility audit, the foundation sprint and monthly monitoring, testing 25 to 40 real buyer questions across all four platforms.
Record every run
For each question and platform we save the date and time, the exact wording, the location and signed-in state, the full answer, the businesses named in order and the sources linked.
Repeat where it matters
Answers can change between runs. For questions that decide enquiries we repeat the run and report when the answers differed, instead of quoting the most flattering one.
Report what was recorded
Results are stated as counts of recorded runs, for example "named in 9 of 30 questions on the dates shown". They are not a ranking, a share of voice for all users or a forecast.
Limits of the method
- Assistants answer differently for different people, places and moments, and platforms change over time. A recording is evidence of that run only.
- A set of 25 to 40 questions describes one business and one market. It is not a representative sample of everything buyers ask.
- Mentions, cited pages, search demand and counted enquiries are reported separately. A mention is not an enquiry.
- Review counts are one input among many and not a universal AI ranking signal, so they are never presented as a score for visibility.
- The readiness checker is a heuristic readiness checklist, not a measured ranking, and not a probability of being recommended.
Find out who AI recommends in your market.
Send your website and the market you want to win. Get a clear view of the competition, your biggest gaps and the next move worth making.