# ChatGPT vs Perplexity vs Gemini: how each one chooses which B2B brands to name

> None of the three publishes how it picks brands. Here is what their docs say, what citation studies show, and what B2B teams should do with it.

Source: https://www.fame.so/post/chatgpt-vs-perplexity-vs-gemini-b2b  ·  Fame Team  ·  November 3, 2026

## Key takeaways

- No engine publishes a formula for choosing brands, so all outside findings are observed patterns.
- Perplexity searches live for each answer, Gemini can ground in Google Search, and ChatGPT search depends on OAI-SearchBot access.
- Studies disagree and most use consumer prompts or come from vendors.
- Measure your own buyer questions on each engine, since citation mixes can shift within weeks.

None of the three engines publishes how it chooses brands. Their documentation explains how they find sources, and outside studies show patterns: ChatGPT's source mix has shifted sharply, Perplexity searches live and cites widely, and Gemini can ground answers in Google Search. Most of that data is not B2B-specific, so test your own buyer questions on each engine.

## What do the engines themselves say?

The official documentation is thin but worth starting with, because it is the only part that is not inference.

- ChatGPT: OpenAI's [crawler documentation](https://developers.openai.com/api/docs/bots) says OAI-SearchBot surfaces websites in ChatGPT's search features. Sites that opt out will not appear in search answers, though they may still show up as navigational links. GPTBot is a separate crawler for training data.
- Perplexity: its [help center](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) says it searches the internet in real time and that each answer includes numbered citations linking to original sources.
- Gemini: Google's [developer documentation on grounding](https://aistudio.google.com/docs/google-search) says the model decides whether a search would help, generates queries, runs them and cites sources inline. This describes the developer tool, and Google does not say the consumer app behaves identically.

None of these documents explains why one brand gets named over another. Anyone who claims to know the exact formula is guessing.

## What does the outside data show?

Three recent datasets are useful, each with limits.

SourceScaleFindingLimitMeltwater, Sept 20264,000+ prompts, about 6.1M citationsTop domain: NIH on ChatGPT (GPT-5), Yahoo on Perplexity Sonar Pro, YouTube on Gemini 2.5 ProConsumer categories onlyOtterlyAI AI Search IndexAbout 6.1M citations, 17 US industries, July and August 2026Company-owned pages were 53.7% of all citationsVendor study, no B2B splitYext6.8M+ citations across 1.6M responses52.15% of Gemini citations came from brand-owned sites, and 48.73% of ChatGPT's from third-party sitesVendor study, examples lean local and consumer

Meltwater, OtterlyAI and Yext all sell products in this market, so read their work as useful observation rather than neutral science. Each one also asks different questions, which is one reason the numbers do not line up.

## How does ChatGPT seem to choose?

The most striking recent finding is instability. OtterlyAI reports that ChatGPT changed on August 8 and again on August 14, 2026. Its company-owned share rose from 44.38% to 55.28%, news and media fell from 21.3% to 10.4%, and Reddit fell from 4.42% to 0.39% of ChatGPT citations. If those figures hold, advice written in spring about winning ChatGPT through Reddit threads may already be stale.

For a B2B team, this suggests two practical steps. First, make sure OAI-SearchBot can reach your site. Second, make sure your own pages clearly state what you do, who you serve and how you differ, since owned pages took a larger share of ChatGPT citations in that data. This is a reasonable inference and not a promise.

## How does Perplexity seem to choose?

Perplexity's live search means fresh pages have a direct route in, provided they can be found. In Meltwater's September data its leading domains were Yahoo and LinkedIn. LinkedIn's position at 53,947 citations is notable for B2B teams, although the prompts were consumer categories. Yext found niche sources made up 24% of Perplexity's citations on subjective, unbranded queries, the highest of the three engines it compared.

The practical reading is that Perplexity may reward breadth. Your site, specialist industry sources and professional profiles can all count. Test this on your own category, because a domain's strength in consumer prompts may not carry over.

## How does Gemini seem to choose?

Gemini's picture is the most contradictory. Meltwater counted YouTube as its top domain, with Reddit close behind. OtterlyAI's [YouTube study](https://otterly.ai/blog/youtube-ai-citation-study-2026/) found Gemini accounted for just 0.2% of YouTube citations, against 38.7% for Perplexity. The two studies use different prompts and periods, and they may not even be reaching Gemini the same way. We cover this conflict in [our post on Gemini and YouTube](https://www.fame.so/post/gemini-youtube-citations).

Yext's result that brand-owned sites supply more than half of Gemini's citations points to a sensible base. Keep your own pages clear, factual and indexed in Google. Google [says](https://developers.google.com/search/docs/appearance/ai-features) that indexing and snippet eligibility are the requirements for its AI Overviews and AI Mode features. It does not make the same statement about the Gemini app, so treat this as a sensible baseline rather than a documented rule.

## Why do the studies disagree so much?

Several things change the result before any engine has even answered. The prompt set is the largest factor, because questions about fast food and streaming services will surface different sources from questions about procurement software. The counting method matters too: one study may report a domain's share of an engine's citations, while another reports an engine's share of a domain's citations. Those sound alike, but they answer different questions.

Timing and access also matter. Meltwater labels the versions it tested, such as Gemini 2.5 Pro and Perplexity Sonar Pro, and a tested version may not behave exactly like the app your buyer has open. Evertune, which also sells in this market, [notes](https://www.evertune.ai/resources/insights-on-ai/youtube-ascendant-on-googles-llms) that not every prompt triggers a web search and that a cited URL does not necessarily shape the response. That second caveat matters for anyone reporting citations as if they were influence. A brand can also be named without being linked, or linked without being recommended, so record both the mention and the citation when you test.

## What is different about B2B?

We did not find a large public study limited to B2B buying questions, so the honest answer is that most of the data above is borrowed from other categories. Meltwater's prompts were consumer. OtterlyAI's index found news dependence ranged from 32.2% in Technology to 4.4% in AI Data Governance, a useful reminder that industry changes the source mix.

B2B questions also tend to be narrower than "best streaming service." A buyer might ask for a compliance vendor for mid-sized manufacturers. The set of pages discussing that exact question may be small, so one well-built page or one good third-party list could matter more than it would in a crowded consumer category. Test that assumption on your own queries before acting on it.

## What should a B2B team do?

- Write 20 to 50 questions buyers really ask, using sales calls and forms, and include questions that do not mention your brand.
- Run each question several times on each engine, because answers vary. Record who is named and which sources are cited.
- Check crawler access in robots.txt, including OAI-SearchBot.
- Build or improve pages for the questions where a competitor's page is being cited and you have nothing equivalent.
- Look at the third-party pages cited in your category. Those are the outreach list, and the engines have effectively written it for you.
- Repeat weekly or monthly. The August ChatGPT shift shows that a one-time audit can age quickly.

The free [share of answers calculator](https://www.fame.so/tools/share-of-answers-calculator) can help you turn those records into a simple figure.

Fame publishes this article and sells [Fame Search](https://www.fame.so/search), a $2,000 a month service that does this tracking weekly across ChatGPT, Perplexity, Claude and Gemini. On Fame's own site, between 38% and 76% of tracked AI answers cited fame.so as of October 30, 2026, measured since June 1. That is our own site and not a client result, and Fame has no client results to report yet.

## FAQ

**Which engine should a B2B company prioritize?**

Find out which ones your buyers use, then measure all of them if you can. The source mix differs enough between engines that a result on one tells you little about another.

**Do these engines use the same sources?**

No. Studies show different leading domains on each engine, although the figures depend heavily on the prompts tested and many come from vendors.

**Can I find out exactly why ChatGPT named a competitor?**

Not exactly. You can see which sources an answer cites when it cites them, and that is the best available clue, but no engine publishes its selection logic.

**How often do the patterns change?**

Sometimes within days. OtterlyAI recorded two changes in ChatGPT's behavior within a week in August 2026, which is why repeated measurement beats a one-time audit.