Key takeaways
- There is no ChatGPT ranking position and no way to pay for an organic citation. OpenAI documents its crawlers but not how sources are chosen, so treat any exact ranking-factor percentage as unverified.
- Four OpenAI user agents do different jobs. OAI-SearchBot controls search eligibility, GPTBot controls training use, ChatGPT-User handles user-initiated visits, and OAI-AdsBot only visits pages submitted as ads.
- The controllable sequence is access, then readable pages, then answer-first writing, then third-party mentions, then measurement with the utm_source=chatgpt.com referral tag and repeated prompt sampling.
- Mention rate across repeated prompts is the metric to track. A single run of a nondeterministic model tells you almost nothing, and answers differ between engines.
In this post9 sections
- Can you actually rank in ChatGPT?
- How does ChatGPT decide which sites to mention?
- Step 1: Can ChatGPT’s search crawler reach your pages?
- Step 2: Can the page be read without running JavaScript?
- Step 3: Does the first paragraph under each heading answer the question?
- Step 4: Are you mentioned on the pages ChatGPT already cites?
- Step 5: How do you measure ChatGPT visibility?
- What does not work, or is unproven?
- What does a 30-day plan look like?
Can you actually rank in ChatGPT?
You cannot rank in ChatGPT the way you rank in Google. There is no results page with ten positions and no place to buy an organic one. ChatGPT search answers cite a few web pages, and OpenAI does not publish how it picks them. What you can influence is eligibility, findability and whether your page is worth citing.
That changes the goal. Chasing position one makes no sense, so track how often your brand or URL appears across repeated prompts. Everything below follows from that: get crawled, get understood, get mentioned where ChatGPT already looks, then measure the mention rate.
How does ChatGPT decide which sites to mention?
ChatGPT draws on two things: what its model learned in training, and pages it fetches live when it searches. OpenAI documents the crawlers behind the second path but not a ranking formula. Studies from SEO vendors describe patterns, but their methods differ and none is an OpenAI statement.
OpenAI’s crawler documentation lists four user agents, and each setting is independent:
| User agent | What OpenAI says it is for | What robots.txt does |
|---|---|---|
| OAI-SearchBot | Search. Surfaces sites in ChatGPT's search features. | Opted-out sites are not shown in ChatGPT search answers, though they can still appear as navigational links. |
| GPTBot | Crawling content that may be used to train generative AI foundation models. | Disallowing it signals the content should not be used for training. |
| ChatGPT-User | Certain user-initiated actions, such as visiting a page when a user asks, and GPT Actions. | Because the user starts the action, robots.txt rules may not apply. It is not used to decide search eligibility. |
| OAI-AdsBot | Validating pages submitted as ads on ChatGPT. | Visits only pages submitted as ads. |
The practical point: search eligibility sits with OAI-SearchBot, and that is the agent to allow if you want to be cited.
Step 1: Can ChatGPT’s search crawler reach your pages?
Allow OAI-SearchBot in robots.txt and confirm your firewall or CDN is not blocking it. OpenAI says a robots.txt change can take about 24 hours to take effect for search. Access is a gate: if the crawler cannot fetch a page, nothing else on this list matters.
A minimal robots.txt that allows search and makes a separate decision on training looks like this:
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /The GPTBot line is a training-data policy, not a visibility setting, so choose it for your own reasons. OpenAI’s publishers FAQ also says a noindex tag only works if the crawler can fetch the page to read it, so do not block a page and rely on noindex at the same time. OpenAI publishes IP ranges for its agents (for example openai.com/searchbot.json), which helps when you need to allow-list them at a firewall.
Step 2: Can the page be read without running JavaScript?
Put the facts you want cited in the server-rendered HTML. OpenAI’s documentation does not say whether its crawlers execute JavaScript, so the safe assumption is that content injected by scripts may be missed. Prices, definitions, tables and answers should be in the HTML a plain request returns.
Test it: fetch the page without a browser (a plain curl or “view source”) and check the key paragraph is there. If your site is client-rendered, a pre-rendering layer is one fix. Two tools in our directory sell that: Alli AI serves pre-rendered, AI-readable pages to AI crawlers, and Scrunch AI serves separate page versions to AI agents.
Step 3: Does the first paragraph under each heading answer the question?
Write each section so its first 40 to 60 words answer the heading on their own. A model lifting that paragraph out of context should still get a complete answer. Add specifics after it: figures, dates, names, a table. Pages built this way are easiest to quote, and they read well for humans too.
Compare two ways to open a section titled “How much does X cost?”:
- Weak: “Pricing is a topic many buyers ask about, and it depends on several factors that we will explore below.”
- Answer-first: “X costs 49 USD a month for one site on annual billing and 69 USD month to month. The plan caps 20,000 credits, and there is no free tier.”
The second can be cited as written. In a manual capture of eight “best AI ___” queries on 29 October 2026 (187 URLs; partial for Perplexity and Gemini, ChatGPT not captured), Bing and Perplexity cited listicles for 81% and 94% of the URLs captured. Of 16 winning listicles reviewed, 14 had a comparison table and nearly all showed per-item pricing and a 2026 date. Treat that as a sample, not a rule, but it matches the answer-first format.
Step 4: Are you mentioned on the pages ChatGPT already cites?
Find the pages that ChatGPT cites for your target prompts and get included in them, because AI answers lean on third-party lists, reviews and discussions as much as on your own site. Your own page is one candidate. A mention across several independent pages is a stronger signal than one well-optimised URL.
A workable routine:
- Write 10 to 20 prompts a buyer would type, such as “best AI SEO tools for a small agency”.
- Run each in ChatGPT with search on and open every cited link.
- List which sites keep appearing: listicles, comparison pages, community threads, reference sites.
- Pitch an accurate, specific inclusion to the editors of those pages, or take part honestly in the threads. Do not astroturf.
Vendor studies report that community and reference sites are among the most-cited sources, but their datasets and methods differ, so verify against your own prompts.
Step 5: How do you measure ChatGPT visibility?
Track two things: referral visits and mention rate. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral links in search results, so in GA4 you can filter session source for chatgpt.com. Visits miss mentions that never produce a click, so also sample prompts and record how often you appear.
Because model answers vary from run to run, one sample means little. Run each prompt several times, record whether your brand or URL appears, and report a rate over the whole set. Manual sampling works for a first month. After that, a tracker automates the sampling and covers more engines. Our guide to the best AI visibility and GEO tools compares the options by price, prompt limits and free access, and the best AI SEO tools guide covers suites that include tracking.
What does not work, or is unproven?
Guaranteed ChatGPT rankings, “AI SEO” schemes that promise a position, and exact ranking-factor percentages are all unsupported. OpenAI publishes no selection formula, so any breakdown of how much authority or schema counts is a vendor estimate. Two common claims are worth testing before you spend time on them.
- llms.txt: OpenAI’s crawler documentation does not mention it. It costs little to publish, but do not expect it to move citations.
- Schema markup: some vendor guides say it helps. We found no controlled public study showing it changes ChatGPT citation rates. Keep it for Google and clarity.
- Paid placement: ChatGPT ads are a separate product, and OpenAI documents a separate agent for them. Nothing in its crawler documentation describes paying for an organic citation.
What does a 30-day plan look like?
Spend week one on access, week two on page structure, week three on third-party mentions and week four on measurement and review. The order matters because each step depends on the one before it. Skipping access or rendering wastes the writing work that follows.
| Week | Do | Done when |
|---|---|---|
| 1 | Allow OAI-SearchBot, check the firewall, decide GPTBot, confirm key facts are in server-rendered HTML | A plain fetch of your top pages shows the answers; robots.txt has been live for more than 24 hours |
| 2 | Rewrite your 5 to 10 highest-value pages answer-first, add tables, prices and visible update dates | Every H2 opens with a self-contained 40 to 60 word answer |
| 3 | Run the prompt set, list the cited third-party pages, pitch inclusion on 5 to 10 of them | You have a target list and outreach sent |
| 4 | Set up the chatgpt.com filter in GA4, rerun each prompt several times, record the mention rate | You have a baseline mention rate to compare next month |
Repeat the sampling monthly. A change in mention rate after a page rewrite is the only evidence you will get that the work moved anything.
Frequently asked questions
Can you pay to rank in ChatGPT?
Not for organic answers. OpenAI documents no placement fee for search citations. It does list a separate agent, OAI-AdsBot, that visits only pages submitted as ads on ChatGPT, which is a different product from organic citations. Be wary of any vendor promising guaranteed ChatGPT rankings.
Should I block GPTBot to rank better in ChatGPT?
Blocking GPTBot does not help or hurt search visibility. OpenAI says disallowing GPTBot means content should not be used for training, while OAI-SearchBot controls search results. Decide GPTBot as a training-data policy and keep OAI-SearchBot allowed if you want to appear in answers.
How long does a robots.txt change take to affect ChatGPT search?
OpenAI's crawler documentation says it can take about 24 hours from a robots.txt update for its systems to adjust for OAI-SearchBot. Content changes on a page you already allow follow your normal crawl timing, which OpenAI does not publish, so recheck after a few days.
How do I see traffic from ChatGPT in Google Analytics?
OpenAI says ChatGPT adds utm_source=chatgpt.com to referral links in search results. In GA4, open the traffic acquisition report and filter session source for chatgpt.com. This counts clicks only, not mentions that never produced a visit, so pair it with prompt tracking.
Does schema markup or llms.txt help me get cited by ChatGPT?
OpenAI's crawler documentation does not mention either. Vendor guides claim schema helps, but we found no controlled public study showing it changes citation rates. Add schema for Google and clarity, and treat llms.txt as optional rather than a ChatGPT ranking lever.
How do I know if ChatGPT mentions my brand?
Run a fixed set of buyer prompts several times each and record whether your brand or URL appears. Dedicated trackers automate this daily across engines. Because answers vary run to run, report a mention rate over many samples rather than a single result.