AI assistants answer buying questions from two places: what the model learned in training, and pages it finds with a live web search. For questions about products and services, the search usually decides. The brands that get named are the ones that show up, clearly described, on the pages AI finds: your own site, and above all the lists, reviews, comparisons and forums that talk about you.
Every answer comes from two places
A model has a memory: what it learned from the web during training, months or years before your buyer asks. On its own that memory goes stale, so ChatGPT, Gemini, Perplexity, Claude and Google's AI Overviews and AI Mode can all search the web when a question needs current facts. Buying questions almost always do.
- A buyer asks“Best CRM for a small sales team?”
- AI searches the webSeveral short searches at once
- AI reads the resultsA handful of pages, not all of them
- AI answersNames a few brands and links some sources
Google calls the searching step query fan-out: the AI splits one question into several related searches, reads what comes back, and writes one answer from it (Google). The other AI platforms work in a similar way, and ChatGPT and Claude show the searches they ran.
AI searches like a buyer in a hurry
The searches behind an answer are short and commercial. On a question about project management tools for agencies, Claude searched for “best project management software for agencies 2026”. ChatGPT's searches look the same.
So the pages AI reads first are this year's “best of” lists, comparisons and reviews. If your brand isn't on those pages, it's rarely in the answer, however good your own site is.
Being cited isn't the same as being named
You might expect the brands whose websites AI cites to be the brands it recommends. In the answers we collected, that often wasn't the case.
| Brand | Named in the answer | Own website cited |
|---|---|---|
| HubSpot | 81% | 36% |
| Salesforce | 49% | 37% |
| Zoho CRM | 57% | 24% |
| Hoka | 56% | 1% |
| Nike | 38% | 22% |
salesforce.com was cited slightly more often than hubspot.com, yet HubSpot was named far more. Hoka was named in more than half of running shoe answers while its own site was almost never cited. Your own site tells AI what you do. What other sites say about you decides whether AI picks you.
Each AI reads different sites
The six AI platforms don't read the same web. Google's AI Overviews leaned on Reddit, YouTube and forums. ChatGPT cited brands' own sites more than any other AI. Perplexity and Gemini favoured review and comparison sites.
- ChatGPT321454
- Gemini2269
- Perplexity142462
- AI Overviews123050
- AI Mode21141847
- Claude1580
Share of all the sites each AI cited, by kind. “Everything else” is news, blogs, directories and other sites. From our research, November 2026.
Being first on one platform tells you little about the others. That's why tracking has to cover each AI your buyers use: see how to track AI visibility.
Early questions name fewer brands
Answers to early, learning questions (“how do I choose a CRM?”) named 2.0 brands on average. Comparing questions (“best CRM for small business”) named 3.3, and choosing questions 2.4.
- Learning2.0
- Comparing3.3
- Choosing2.4
From our research, November 2026.
If buyers meet you only when they're comparing, a competitor may already have their attention. The brands named in early answers get there first.
What makes AI more likely to name you
- Be on the lists AI readsThis year's “best of” pages, comparisons and roundups for your category.
- Earn reviews where AI looksThe review sites AI cites for your industry, with recent, specific reviews.
- Say clearly what you areOne plain sentence on what you do and who it's for, on your homepage and key pages.
- Keep facts currentPrices, features and dates that match everywhere they appear.
- Answer real questionsA page for each question buyers ask, not one page for everything.
- Let AI crawlers inSearch crawlers allowed in robots.txt and at your firewall.
Do
- Find the pages AI cites for your questions and work to be on them
- Describe your product the same way on your site, listings and profiles
- Publish comparisons and pricing pages buyers can trust
Don't
- Rely on your own site alone
- Stuff pages with keywords or invented claims
- Judge it by one answer on one AI
Where to start
- 1
Find the questions
List 20 to 50 questions buyers ask before they choose, none of them naming your brand.
- 2
See who gets named
Ask each question on the AI platforms your buyers use and note who's named, and in what order.
- 3
Find the pages AI cites
Note the sites cited in each answer, above all the ones that name competitors and not you.
- 4
Fix the biggest gaps
Get onto those pages, publish pages for the questions where you're missing, and ask customers for reviews.
- 5
Check again every week
Answers change. Watch the trend across weeks, not a single answer.
Your checklist
0 of 6 doneTicks are saved in this browser only.
Questions people ask
How does ChatGPT decide which brands to recommend?
For buying questions ChatGPT usually searches the web first, reads a handful of pages and names the brands those pages talk about most clearly. Being on current “best of” lists, comparisons and review sites matters most. See how to get recommended by ChatGPT.
Is AI visibility the same as SEO?
It builds on it. Google says its AI features use the same Search index and best practices. The difference is what you measure (whether you're named, not where a page ranks) and where you need to be (the pages AI reads, not just your own).
Can I pay to be recommended by AI?
Not in the answers themselves. What you can do is be on the pages AI reads, with a clear description of what you do.
Why does AI recommend my competitor and not me?
Usually because your competitor appears on more of the pages AI finds for that question: lists, reviews, comparisons and forums. Look at the sites cited in the answer to see which ones.
Do all AI platforms recommend the same brands?
No. In our research, every AI agreed on the first brand to name on only 21% of questions, and each one read different kinds of sites.
Sources
Checked on November 11, 2026. Numbers about AI answers come from our research; everything else from the companies' own pages.