fm.

How to rank in ChatGPT: what gets a brand named and cited

ChatGPT has no second page. A brand is named in the answer, cited as a source or left out. What decides which, the technical floor that keeps you eligible and the 30-day plan I'd run.

Key takeaways

  • ChatGPT names brands from 2 places: what the model learned in training and what it finds when it searches. You can move the second one this month.
  • The floor is technical: OAI-SearchBot allowed, pages indexed in Bing, answers in server-rendered HTML.
  • Above the floor, citations go to pages that answer first with specific numbers, and to the lists, reviews and comparison pages ChatGPT already cites for your category.
  • Judge it on a fixed prompt panel and on pipeline. At Pipe17, 32% of new-business wins came through AI assistants, revenue the CRM had credited to paid search.

01The short answer

To rank in ChatGPT, let OpenAI's search crawler and Bing reach your pages, answer buyer questions in the first 2 sentences with specific numbers and get your brand named on the third-party pages ChatGPT already cites for your category. ChatGPT has no position 1, so ranking means being named in the answer or cited as a source when a buyer asks, and you measure it with a fixed panel of buyer prompts.

People call this ChatGPT SEO, AEO or GEO. I cover how AEO, GEO and SEO differ separately, and the full program is in my AEO playbook. This note goes deeper on one question: what makes ChatGPT name a brand.

At Commas I analyzed 10,560 ChatGPT and Google AI Overview answers in Profound. Commas earned 8.4% of citations against 21% to 33% for peers, and that gap became 221 page opportunities and a 35-topic roadmap.

02How ChatGPT decides what to name

Every answer draws on 2 kinds of knowledge.

  • Trained knowledge. What the model learned from the web before its training cutoff. It decides whether ChatGPT knows your brand, and which category it files you under, when it answers without searching. It changes slowly, model by model.
  • Live search. OpenAI says ChatGPT may search the web automatically when a question would benefit from current information. It can send rewritten versions of the prompt to partner search providers, and the answer can cite the sources it used. OpenAI's help center names Microsoft among those providers, and OpenAI runs its own search crawler, OAI-SearchBot.

Most buying questions need current information, like prices and this year's options, which makes a search likely. A page can be cited there only if ChatGPT can fetch it and finds it for the rewritten queries. Then it has to state the answer plainly enough to quote.

Answers also vary by user. With memory on, ChatGPT may use saved memories when it rewrites a search query, and it can use an approximate location from the user's IP address, so one screenshot proves nothing. There is no paid shortcut either. OpenAI began testing ads in ChatGPT in 2026 and says ads are labeled, kept separate and do not influence answers.

03The technical floor

None of this wins a citation on its own. It keeps you eligible, so I check it first.

  • Allow OAI-SearchBot. OpenAI's crawler documentation says sites that opt out of it are not shown in ChatGPT search answers, though they can still appear as plain links. Robots.txt changes take about 24 hours to apply.
  • Leave ChatGPT-User open. It fetches pages when a user's request needs them. OpenAI says robots.txt rules may not apply to it, so check the firewall.
  • Decide GPTBot separately. It's the training crawler, and OpenAI treats each setting independently, so blocking it doesn't remove you from search. On a marketing site I allow it, because trained knowledge is the slow half of being named.
  • Check the CDN and firewall. Bot-management rules can turn AI crawlers away even when robots.txt allows them. Server logs should show OAI-SearchBot getting 200 responses.
  • Get indexed in Bing. Verify the site in Bing Webmaster Tools, submit the sitemap and send an IndexNow ping on every publish and update. IndexNow shares each URL with every participating engine, and Microsoft recommends it so AI systems hear about changes.
  • Serve the answer in HTML. In a December 2024 analysis of AI crawler traffic on its network, Vercel found that none of the major AI crawlers, OpenAI's included, rendered JavaScript. If your pricing table loads client-side, assume ChatGPT can't read it.
  • Add structured data. Organization, Product, Article and FAQPage markup won't earn a citation alone. It removes guessing about who publishes the page and what it describes.
  • Treat llms.txt as optional. Google has said it doesn't use the file and OpenAI hasn't said it does. I keep one because it takes 10 minutes, and I don't count on it.

04The content that gets cited

Once a page is eligible, it has to be the easiest one to quote.

  • Answer first. The buyer's question as the heading, the answer in the 2 sentences under it. Models lift whole sentences, so write ones that stand alone.
  • Specific numbers. Prices, limits, dates and named integrations. “Fast setup” gives a model nothing to repeat. A setup time in days does.
  • Comparison tables. The same attributes on every row, including rows where a competitor wins. Buyers close to a decision ask comparison questions.
  • Entity clarity. One name, one category line and the same facts on your site, profiles and listings. If ChatGPT files you in the wrong category, start with your About page.
  • Freshness. A visible last-updated date and current figures. Stale dates get passed over.

Microsoft's guidance for getting cited in its AI answers says much the same: clear headings, tables, FAQ sections, claims backed by data and content kept current.

05Third-party presence

For category and comparison prompts, the pages that decide the answer are often ones you don't own. Cited sources cluster: in most categories a handful of review sites, listicles, documentation pages, community threads and comparison pages carry most of the citations. I build the outreach list from those citations.

  • Lists. Ask the authors of the best-of articles cited for your prompts to add you, with the facts that make you a fit. Skip lists nobody cites.
  • Reviews. Ask customers to review you on the cited review sites and to name the job you did. A named use case is easier to quote than a star rating.
  • Communities. Answer real questions in cited threads as yourself, affiliation stated. Planted praise tends to get removed.
  • Comparisons. Publish honest versus and alternatives pages, and ask third-party comparisons to fix facts they get wrong.

My decision rule: a page cited across many prompts and engines is worth more than 10 new mentions nobody cites.

06The ranking factors in one table

Ranking factorWhat to doHow to check
Crawler accessAllow OAI-SearchBot and ChatGPT-User in robots.txt and the firewall.Logs show OpenAI's crawlers getting 200 responses.
Search indexBing Webmaster Tools, a sitemap and IndexNow on every change.Bing's URL inspection shows the page indexed.
Readable HTMLServer-render answers, tables and prices.The answer is in the raw page source.
Answer-first pagesA 2-sentence answer under each question heading.The opening lines make sense quoted alone.
Specific factsNumbers, dates, named entities, comparison tables.Count the claims a model could repeat word for word.
Entity clarityOne name and description everywhere, Organization markup with sameAs.ChatGPT describes you the way your About page does.
FreshnessA visible updated date, current figures, a matching dateModified.Page and markup dates agree.
Third-party presenceGet onto the cited lists, reviews, threads and comparisons.The share of cited third-party pages that name you.
MeasurementA fixed panel, chatgpt.com referral tracking, a self-reported source field.Citation share and AI-sourced pipeline, monthly.

07A 30-day plan

  1. Days 1 to 3: open the doors.Fix robots.txt and firewall rules for OpenAI's crawlers and set up Bing Webmaster Tools, the sitemap and IndexNow. Confirm your top pages show their answers in raw HTML.
  2. Days 4 to 7: run the baseline.Write 50 to 100 buyer prompts across category, comparison, problem and brand intents. Run them in ChatGPT with search on and record mentions, position and every cited URL.
  3. Days 8 to 14: fix your own pages.Rewrite the 5 pages closest to a buying decision: answer first, a comparison table, specific numbers, a visible date. Make the About page state the facts you want repeated.
  4. Days 15 to 21: work the cited sources.Rank third-party pages by how often they're cited, then pitch the top lists, ask for reviews on the cited sites and correct wrong comparisons.
  5. Days 22 to 30: measure and queue.Rerun the same panel against the baseline, add AI assistants to your self-reported source field and queue the next 10 pages.

Technical fixes show first. Pages need recrawling and third-party edits take weeks, so I judge content and outreach on the panel 1 to 2 months later.

08How to check whether it's working

  • A fixed prompt panel. Same prompts, same schedule, read by topic. The core number is citation share: your citations divided by all citations in the panel. At Commas, 8.4% against 21% to 33% set the page list.
  • Referral sessions from chatgpt.com. OpenAI says ChatGPT adds utm_source=chatgpt.com to the links it sends. Trend those visits monthly and follow them to conversions.
  • Self-reported attribution. Add AI assistants to the how-did-you-hear field. At Pipe17, rebuilding attribution around what buyers said showed that 32% of new-business wins came through AI assistants.

Bing Webmaster Tools' AI Performance report adds a proxy: citations and grounding queries from Copilot, Bing's AI summaries and some partners Microsoft doesn't name. The full metric set and the GA4 setup are in how to measure AI visibility.

09Mistakes to avoid

  • Chasing one prompt. One prompt is a sample of one, and the answer shifts by user and run. Judge the topic across all its prompts.
  • Changing the prompts weekly. The trend stops meaning anything. Add new prompts as a separate cohort.
  • Keyword-stuffed pages. Repetition gives a model nothing new to quote. One clear answer with a number beats 5 restatements.
  • Hiding the brand behind vague copy. If your homepage says “the platform for modern teams,” ChatGPT can't tell what you sell or to whom. Name the category and the buyer in the first line.

FAQQuestions people ask

How do I rank higher in ChatGPT?
Make sure OpenAI's search crawler and Bing can reach your pages and put a 2-sentence answer with specific numbers at the top of the pages buyers need. Then get named on the third-party lists, reviews and comparison pages ChatGPT already cites for your category and track it on a fixed panel of buyer prompts.
How do I show up in ChatGPT search?
To show up and rank in ChatGPT search, allow OAI-SearchBot in robots.txt and your firewall, serve your content as server-rendered HTML and get indexed in Bing through Bing Webmaster Tools and IndexNow. OpenAI says sites that opt out of OAI-SearchBot are not shown in ChatGPT search answers.
Is ChatGPT SEO different from regular SEO?
The foundation is shared: crawlable, indexed pages with clear answers. The difference is the result. ChatGPT gives a short answer with a few named brands and cited sources, so you measure citation share across buyer prompts and spend more time on the third-party pages it cites.
How do you get cited by AI?
Publish pages that answer the question in the first 2 sentences, back the answer with specific numbers and tables, keep them current and make them readable without JavaScript. Then get named on the third-party pages AI engines already cite in your category.
How do I get recommended by ChatGPT?
State who you're for and what you do in the same words on your site and every listing, so ChatGPT can match you to the buyer's situation. When it searches, its recommendations come with sources, so get onto the reviews, lists and comparison pages it cites for that use case.