Your brand can appear on Perplexity but not ChatGPT because each engine runs its own web search, reads a different set of websites and writes its own shortlist of brands. In our study of 480 AI answers to buying questions asked from inside India, ChatGPT and Perplexity shared only 32% of the brands they named for the same question, while two ChatGPT answers to that same question shared 55%.
Shared brands means the brands two answers have in common, divided by all brands either answer named. A gap between engines is normal, and you can measure it. Below: how each engine builds an answer, what our data shows, and a checklist of causes to test.
Why does my brand appear on Perplexity but not ChatGPT?
The two engines read very different parts of the web before they answer. A citation is a link to a source page that an engine shows next to its answer, and citations are the visible trace of where an answer came from.
In English, Perplexity showed 19.6 citations per answer on average. ChatGPT, with web search on, showed 2.5. We asked every question in English and in Hinglish, which is Hindi-English typed in English letters ("India mein oily skin ke liye 500 rupees ke andar konsa face wash kharidna chahiye?"). Across 160 answers each, Perplexity cited 167 different websites and ChatGPT cited 76.
The mix of websites differs too:
- Perplexity cited reddit.com in 101 of its 160 answers. ChatGPT cited it in 3.
- Perplexity cited youtube.com in 88 answers. ChatGPT cited it in 2.
- Perplexity cited amazon.in in 79 answers and flipkart.com in 70. ChatGPT cited amazon.in in none and flipkart.com in 4.
When we matched the two engines on the same question in the same language, ChatGPT had cited only 12% of the websites Perplexity used (43 of 358 question and website pairs).
The brand lists split just as sharply. For each question we pooled the 8 runs per engine and language. Perplexity named 163 brand and question combinations at least once. 70 of them (43%) never appeared in any of ChatGPT's 8 answers to the same question in the same language. The gap runs the other way as well: 64 of ChatGPT's 157 combinations (41%) never appeared on Perplexity.
If buyers discuss your brand mostly in Reddit threads, YouTube reviews and marketplace listings, Perplexity's source mix gives it more ways to find you. ChatGPT cites far fewer pages per answer, so a handful of articles and brand sites shape its shortlist. Citations show what an engine displayed. They do not prove why it named a brand, so use them as the first place to look.
How each AI engine builds an answer
ChatGPT. OpenAI's help page on searching the web with ChatGPT says ChatGPT may search the web on its own when a question needs current information. It often rewrites your question into one or more shorter search queries, sends them to search partners, and can use an approximate location based on your IP address. If memory is on, saved memories can shape those queries too. The same page says a site must allow OAI-SearchBot, the bot OpenAI uses to fetch pages for search, to be eligible for ChatGPT search results. OpenAI's crawler documentation adds that sites which disallow it "will not be shown in ChatGPT search answers". GPTBot, OpenAI's training crawler, is a separate setting.
Perplexity. Perplexity's help page on how Perplexity works says it searches the internet in real time, summarises what it finds, and adds citations that link to the original sources. Its crawler guide lists two bots. PerplexityBot surfaces and links websites in Perplexity search results. Perplexity-User visits pages when a person asks a question, and the guide says it generally ignores robots.txt, the file that tells bots which pages they may fetch. In our data Perplexity was the most source-heavy engine and also the most repeatable: two runs of the same question shared 68% of their brands and 85% of their cited websites.
Gemini. Gemini is Google's AI assistant at gemini.google.com. Google's developer guide to grounding with Google Search says a Gemini model first decides whether a Google Search would improve the answer, then writes one or more search queries, runs them and builds its reply from the results with links to its sources. That guide covers the Gemini API. Google's help page on sources in Gemini Apps says the app may show links to sources, including public websites, below or within a response. In our captures it attached sources to 72 of 80 English answers, with 9.4 citations per answer, and wrote the longest English answers (4,609 characters on average). Its brand lists sat closer to ChatGPT's (38% shared) than to Perplexity's (33%).
Google AI Overviews. An AI Overview is the AI-written summary Google shows above some search results. Google's guide to AI features in Search says a page must be indexed and eligible to show in Google Search with a snippet to appear as a supporting link, with no extra technical requirements. Our study did not include AI Overviews, so this post has no comparison figures for them.
What our study shows about engine differences
The figures below come from 480 answers: 10 buying questions (5 skincare, 5 fashion), each in English and Hinglish, asked 8 times to each engine from inside India on 14 Aug 2026. "When asked again" compares repeat runs of the same question on one engine. Engine pairs compare runs of the same question in the same language.
| Measure | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Citations per answer, English | 2.5 | 9.4 | 19.6 |
| Citations per answer, Hinglish | 2.6 | 5.2 | 18.8 |
| Different websites cited, 160 answers | 76 | 97 | 167 |
| Brands named per answer | 4.3 | 5.5 | 5.3 |
| Brands shared when asked again | 55% | 51% | 68% |
| Same first brand when asked again | 60% | 63% | 69% |
| Extra brand change, English vs Hinglish | 2.2 points | 7.8 points | 23.4 points |
The last row shows how much less brand lists matched between English and Hinglish versions of a question than between two identical runs.
| Engine pair | Brands shared, same question | Same first brand |
|---|---|---|
| ChatGPT and Gemini | 38% | 39% |
| ChatGPT and Perplexity | 32% | 18% |
| Gemini and Perplexity | 33% | 15% |
What the two tables show:
- Every pair of engines agreed less than any single engine agreed with itself. The lowest self-match was Gemini at 51%. The highest match between two engines was 38%.
- ChatGPT and Perplexity picked the same first brand in 18% of answer pairs. Each engine repeated its own first brand in 60% to 69% of reruns. Being named first on one engine tells you little about the other.
- Sources barely overlapped across engines. Two answers to the same question from different engines had 4% to 5% of their cited websites in common, against 33% to 85% for reruns on one engine.
Single brands show the split clearly. These are English answers that named the brand, out of 80 per engine:
| Brand | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Plum | 9 | 15 | 23 |
| Taneira | 0 | 1 | 12 |
| La Roche-Posay | 0 | 0 | 11 |
| The Derma Co | 20 | 30 | 2 |
| Minimalist | 29 | 36 | 18 |
Taneira and La Roche-Posay match the question in this post's title: regular on Perplexity, absent from ChatGPT. The Derma Co shows the reverse, named in 20 ChatGPT answers and 2 Perplexity answers. The full study is on the English vs Hinglish AI shopping study page.
Method and limits
The answers come from Depra's English vs Hinglish study, collected on 14 Aug 2026 in one 57-minute window. ChatGPT answers were captured from the chatgpt.com web app with search on, Gemini answers from the gemini.google.com web app, and Perplexity answers from Sonar, the answer model Perplexity sells through its API, with live web search. All requests were logged out and located in India. The rerun, language and citations-per-answer figures are from the published study. The engine-pair figures, pooled brand gaps, website counts and website overlap are our re-analysis of the same 480 stored answers, run on 15 Sep 2026 with the study's fixed brand list and matching rules. Limits: 10 questions in two categories, one country, one day, and no logged-in history or memory. Brand counts cover all 10 questions, so a skincare brand could only be named in the 40 skincare answers. Brands missing from the study's brand list were not counted. Citations are the links engines displayed, which do not prove why a brand was named. Depra sells an AI visibility tracker, so read the numbers with that interest in mind.
Why one AI model recommends your brand and another does not
You can test each cause below yourself.
- The engine's crawler is blocked. If your robots.txt blocks OAI-SearchBot, OpenAI says your pages will not show in ChatGPT search answers, even when PerplexityBot is allowed. Check each bot by name. Check your firewall and CDN (the service that delivers your site) too: OpenAI's help page asks sites to allow traffic from its published search bot IP addresses.
- The engine cites websites that leave you out. Open 5 to 10 answers on each engine and list the cited pages. If Perplexity cites Reddit threads and marketplace listings where you appear, and ChatGPT cites two roundup articles that skip you, those two articles are your target.
- The question's language. In our study, Perplexity's brand lists matched 23.4 points less between English and Hinglish versions of a question than between identical runs. ChatGPT's matched 2.2 points less. A brand that shoppers discuss in Hinglish can appear for Hinglish questions on one engine and not the other.
- Location. ChatGPT can use an approximate location from your IP address, and a VPN can change it. Our Perplexity requests set India as the search country. A check run from India and one run through a foreign VPN can return different brands.
- One check is not a result. Each engine changes its own answer between runs. Two ChatGPT answers to the identical question shared 55% of their brands. A single check where you are missing on ChatGPT and present on Perplexity can flip on the next try, so ask each question several times before you decide anything.
- Personal context. OpenAI's help page says saved memories can shape ChatGPT's search queries. Test logged out, or with memory turned off, so your own history does not tilt the result.
How to find out which engine is the problem
You can run the test by hand. Pick 10 buying questions that do not contain your brand name. Ask each one 5 times on ChatGPT with web search, on Perplexity and on Gemini, logged out. Record the brands named, your position in the list and the cited links. That is 150 answers per round. If you are missing on one engine across most runs and present on another, compare that engine's cited websites first. Our guide to tracking brand mentions in AI answers covers the sheet setup.
Depra automates this check. It asks your buyers' questions to ChatGPT (the web app with search on), Gemini and Google AI Overviews daily, and to Perplexity (Sonar with live web search) weekly, all from inside your country. Every full answer is stored. You see how often each engine names you, which websites each engine cited, and the websites that recommend your competitors but never mention you. Questions that already contain your name are not counted. Each engine has its own page: the ChatGPT visibility tracker, the Perplexity visibility tracker and the Gemini visibility tracker. The AI crawler checker shows which of 41 AI crawlers your robots.txt allows or blocks, with the exact rule for each bot. If ChatGPT is the engine that leaves you out, the guide to why ChatGPT doesn't mention your brand goes deeper.
See which engines name your brand and which skip it. Start 7 days free on any plan, no card.