ChatGPT doesn't publish a ranking algorithm, but its answers reliably favour brands that are easy to access, clearly defined and recommended by many independent sources. Below are 15 signals that consistently influence whether ChatGPT names your brand or cites your pages, grouped into access, understanding, content and consensus — with what to do about each.
A note on certainty: OpenAI hasn't listed ranking factors. These signals follow from how ChatGPT search works (it retrieves pages with OpenAI's crawler and search partners, then writes an answer) and from what we see when tracking prompts. Test them on your own prompts.
How ChatGPT decides, briefly
ChatGPT answers either from trained knowledge or, for current, local, shopping and comparison questions, by searching the web and citing pages. Factors that help retrieval work within weeks; factors that shape trained knowledge take months. See how AI assistants generate answers and how ChatGPT search chooses its sources.
Access factors
1. OAI-SearchBot and ChatGPT-User are allowed
If OpenAI's search crawler or user fetcher is blocked in robots.txt or by your CDN, ChatGPT search can't use your pages. See AI crawlers and robots.txt.
2. Content is in the HTML
OpenAI's crawlers generally don't run JavaScript. Client-rendered content may be invisible. See JavaScript rendering.
3. Broad indexing, including Bing
ChatGPT uses search partners alongside its own crawler; being well indexed in Bing as well as Google widens your chances. See Bing Webmaster Tools and IndexNow.
4. Fast, reliable responses
User-initiated fetches happen in real time. Slow pages and errors get skipped. See page speed for AI fetchers.
Understanding factors
5. A clear, consistent entity
The same name, category and description everywhere — site, LinkedIn, directories, review profiles. See brand consistency and entity signals.
6. Explicit audience and category
"Accounting software for Indian SMBs" matches prompts; "solutions that empower businesses" matches nothing.
7. Accurate public facts
Pricing, features, locations and policies published openly. Hidden facts lead ChatGPT to guess or rely on outdated third-party pages. See fixing AI hallucinations about your brand.
Content factors
8. Answer-first structure
Direct answers in the first sentences of pages and sections. See answer-first writing.
9. Specific, quotable facts
Numbers, names, dates, specs and prices in plain text and tables. The original GEO research found adding statistics, quotations and citations made sources more visible in generated answers. See GEO research on statistics and quotes.
10. Coverage of the sub-questions
ChatGPT often runs several searches per prompt. Pages and clusters that cover pricing, alternatives, use cases and setup get retrieved more. See query fan-out explained.
11. Comparison and alternatives pages
"X vs Y" and "alternatives to X" are frequent prompts. Fair, specific pages win citations. See comparison pages.
12. Freshness on time-sensitive topics
Visible update dates and real updates for prices, lists and product facts. See content freshness.
Consensus factors
13. Reviews
Review platforms are heavily cited for recommendations. Volume, recency and detail all matter. See reviews and AI recommendations.
14. Third-party lists and media
Inclusion in the "best X for Y" roundups and publications ChatGPT cites for your category. See getting included in best-of lists and digital PR for AEO.
15. Community discussion
Genuine mentions in Reddit and Quora threads. See Reddit and AI answers and Quora and AI answers.
Factor summary
| Group | Factor | Speed of impact |
|---|---|---|
| Access | Crawler access, HTML content, indexing, speed | Days to weeks |
| Understanding | Consistent entity, clear category, public facts | Weeks to months |
| Content | Answer-first, specifics, coverage, comparisons, freshness | Weeks |
| Consensus | Reviews, lists, media, communities | One to three months (longer for model knowledge) |
What doesn't work
- Keyword stuffing — the GEO research found it didn't help.
- Hidden instructions to AI in your pages — manipulative and unreliable.
- Fake reviews or Reddit accounts — against platform rules and easily backfire.
- Single screenshots as proof — answers vary run to run. See AI answer variability.
Factors people overrate
- Domain age alone. Older domains often have more coverage, but age itself isn't what ChatGPT reads.
- Word count. Long pages help only if they cover more sub-questions clearly.
- Exact-match keywords. ChatGPT matches meaning; specific facts beat repeated phrases.
How to test these factors
Track a fixed set of buyer prompts in ChatGPT weekly, change one thing at a time (fix crawler access, rewrite a page, win a list placement) and watch mention and citation rates. See testing AEO changes. AEO-Rank's ChatGPT visibility tracking runs these prompts for you and shows position, sentiment and cited sources.
