How to Rank on ChatGPT: Get Named When Buyers Ask

That definition matters because the list keeps moving. SparkToro and Gumshoe had 600 volunteers run 12 prompts 2,961 times across ChatGPT, Claude and Google's AI, and the odds of ChatGPT giving the same list of brands twice were under 1 in 100. The odds of the same order were under 1 in 1,000.

The good news is that most categories are still open. A July 2026 Semrush study of 1,094 US categories found that in 53.7% of them, no brand appeared in even three of five buyer prompts in ChatGPT. Only 15.2% had a clear owner.

This guide covers how to rank on ChatGPT for those unbranded "best X" prompts: how ChatGPT builds its shortlist, a worked example with real arithmetic, and the pages to work on first. Getting your pages linked as sources is a related job, covered in our playbook on getting cited by ChatGPT. Crawler settings live in our ChatGPT SEO guide.

What It Means to Rank on ChatGPT

ChatGPT has no results page and no position one. When a buyer asks for the best tool in your category, ChatGPT writes a short list in prose, and that list changes from run to run. So when founders say they want to rank on ChatGPT, they mean three things at once: being named, being named often, and being named near the top.

A mention is not the same as a citation. A mention is your brand name inside the answer. A citation is a link to your page in the Sources panel. The two often part ways: ChatGPT can recommend a brand without ever opening its website. This guide is about the mention, and about the order of the list.

The three numbers that make up your rank

Because every answer is different, rank is a rate you work out across many runs. Three numbers cover it.

SparkToro's authors make the same call. They suggest tracking how often a brand appears across many prompts and runs, instead of its spot in a single answer. Their data shows why: City of Hope appeared in 69 of 71 answers to one prompt, a 97% visibility rate, even though the lists around it kept changing.

Why unbranded prompts decide whether you rank on ChatGPT

Branded prompts, like "is Plannora any good?", test your reputation. Unbranded prompts, like "best project management tool for a startup," test whether ChatGPT thinks of you at all. That second kind is where new buyers come from.

Unbranded prompts are also broader than they look. In one set of captures, a buyer's complaint that "our meeting notes are a mess" became a search for one named brand's official pricing before ChatGPT read a single page. Prompts asking for "alternatives to" a rival pulled in new names too. If a buyer doesn't name a product, ChatGPT supplies its own.

The Shortlist Loop: How ChatGPT Builds a Recommendation List

OpenAI doesn't publish how it picks brands. But its help pages, plus researchers who read ChatGPT's search traffic, give a clear enough picture to act on. We call it the shortlist loop: five stages inside every answer, and one slower loop around them. Every brand that wants to rank on ChatGPT has to pass through it. The model is ours. The evidence under each stage is not.

Stage 1: Trigger

ChatGPT "may search the web automatically when your question would benefit from current information," according to OpenAI's search help page. Buying questions nearly always qualify, because products and prices change. In one researcher's tests, 10 of 11 requests for a recommendation that named no brand sent ChatGPT to search with brand names it supplied itself. Questions about how something works often skipped search entirely.

Stage 2: Recall

OpenAI says ChatGPT "typically rewrites your query into one or more targeted queries" before it sends them to search providers. For a recommendation prompt, that first query often carries a shortlist. Suganthan Mohanadasan read his own ChatGPT traffic for Search Engine Journal and found that in 21 of 27 conversations, the first search held brands the user never typed.

That first query may be where most brands win or lose their chance to rank on ChatGPT. Brands inside ChatGPT's own query were named in the answer 68.9% of the time. Brands it only fetched along the way were named 2.1% of the time. He is clear that this comes from one account, so treat the numbers as direction, not measurement. Memory can shape this stage too: with memory on, OpenAI says ChatGPT "may use relevant saved memories when rewriting a search query."

Stage 3: Probe

Next come the follow-up searches, or query fan-out. They tend to fall into three kinds:

1. Vendor checks. Searches aimed at each named brand's own site, often with "official" and "pricing" in them. Nectiv's study of about 4,000 prompts found the site: operator in 64% of fan-out queries, and "official" was the second most common word.
2. Roundup and review searches. Queries like "best X 2026" or "X reviews," which pull in listicles and review sites.
3. Opinion checks. A Reddit search about the brands already picked. In one capture, ChatGPT searched Reddit about six vendors it had already named, with what looks like a one-year recency window. Vendor checks in the same call carried 30 days.

Counts disagree: Nectiv measured 7.61 searches per prompt in 2026, up from 2.17, while Mohanadasan's captures fell from 12 to 4 in early August 2026. Plan for the kinds of checks, not a count.

Stage 4: Tally

ChatGPT reads many pages and credits few. In Mohanadasan's data, 3,554 retrieved pages across 57 conversations led to 110 citations, or 3.1%. What seems to carry a brand into the answer is recurrence: showing up across the pages it reads, in a way that fits the buyer's needs.

This matches an older pattern. In February 2025, Seer Interactive's Wil Reynolds noticed that the agencies ChatGPT kept naming were the ones repeated across "best agency" lists and directories, including agencies' own ranked lists. He called it brute force repetition.

Stage 5: Order

Nobody outside OpenAI knows exactly how the list is ordered, and order is the least stable part of the answer. Two clues help. Brands placed in the top third of cited roundups showed up more often in Ahrefs' study of "best" lists. And for products, OpenAI's shopping help says that if a user sets a $30 budget, ChatGPT "will focus more on price." Fit to the buyer's stated need moves you up or down.

The slow loop around the five stages

The five stages run again for every answer. Around them runs a slower loop. The roundups, reviews and threads ChatGPT reads today are part of the public web that future models can learn from. Being on them now is the likeliest way into Stage 2 later. OpenAI hasn't confirmed this part; it's our reading of the evidence below.

That would explain why wide coverage beats raw authority. In a December 2025 study of 75,000 brands, Ahrefs found that branded web mentions correlated 0.664 with ChatGPT visibility. Domain rating correlated just 0.266, and Ahrefs saw almost no link between how many pages a site has and its AI visibility.

Worked Example: Tallying Plannora's Shortlist

Plannora is a made-up project management app for small teams, at plannora.io. Its rivals, their domains and every page below are made up too. The queries are illustrative, modeled on the patterns above. The arithmetic is real, so you can copy the method with your own category.

Plannora wants to rank on ChatGPT for one prompt: "What's the best project management tool for a 12-person startup?"

Step 1: Read the first query

Suppose ChatGPT's first search for that prompt looks like this:

best project management software startups 2026 official pricing Loopcraft Taskmeadow Crewboard

Three rivals are in it. Plannora is not. That single line tells Plannora it has a Stage 2 problem, not just a page problem.

Step 2: List the follow-up searches and the pages they surface

That's 11 pages. Two are vendor pages. The other 9 are third-party pages, and those are the ones to tally.

Step 3: Tally which brands recur

Count how many of the 9 third-party pages name each brand.

Loopcraft is on every page. Plannora is on two: one roundup that lists it eighth, and one Reddit thread where a user mentions it in passing.

Step 4: Run the prompt ten times and score it

Now Plannora runs the prompt ten times in Temporary Chat and writes down each list. The lists vary in length from three brands to five, just as SparkToro saw.

Check the math on one row. Crewboard's positions add up to 25 across 9 runs, so its average is 25 ÷ 9 = 2.8. Every one of those positions is third or better, so its top-3 rate is 9 of 10 runs, or 90%.

Plannora's rank today: a 10% mention rate, an average position of 5.0, and a 0% top-3 rate. The pattern also lines up with the tally. The brands on the most pages, and in the first query, hold the top spots.

Step 5: Turn each gap into a task

Now the arithmetic of the fast loop. If Plannora lands on two more roundups, one review page and one more Reddit thread, it appears on 3 + 1 + 2 = 6 of the 9 pages, or 67%. That ties Crewboard today.

That doesn't promise a top-three spot, because the first-query gap belongs to the slow loop. But it puts Plannora on the pages ChatGPT uses to check its list, and those pages are likely raw material for future shortlists. That is the realistic plan to rank on ChatGPT for this prompt.

Get Onto the Pages ChatGPT Reads for Your Category

Most of the work to rank on ChatGPT happens on other people's sites. Start by running your own category prompt a few times and noting every page in the Sources panel. Then work down this list, in this order.

"Best of" roundups

These are the biggest single source. In Ahrefs' December 2025 study of 750 prompts across software, products and agencies, "best X" blog lists made up 43.8% of all cited page types. Most cited lists were fresh: of 1,100 lists checked for dates, 79.1% had been updated in 2025, and 26% in the two months before the study. And 35% of cited lists sat on low-authority domains, so a small niche blog can still make the cut.

Give each writer a reason to update: current prices, a new feature, one sentence on who you're best for, and a free account to test. Each update is a fresh chance to be included.

Your own comparison pages

Publishing your own "best X" list can work, with limits. In a July 2026 Ahrefs experiment covering 9,886 answers from ChatGPT, Gemini, Perplexity and Copilot, Ahrefs published self-promotional pages for its new conference, Evolve. Of 72 new Evolve mentions, 82% came in answers that cited one of those pages. For the long-known Ahrefs brand, only 6% of new mentions did. And 43% of answers citing the conference pages never named the conference at all. The engine used the page and recommended others.

So an honest comparison helps a new or small brand most. Place yourself only where the facts support it, and match the page to a narrow prompt: the same brand was named in 66.4% of answers to a narrow prompt, but 15.8% for a broader one.

Review sites and directories

Review platforms and directories are roundups with star ratings. Seer's 2025 observation named Clutch among the directories that kept repeating agency names. For products, OpenAI says ChatGPT's review summaries are "based on reviews from public websites." Keep your profiles current on the sites that show up in your sources, and ask happy customers to review you there.

Reddit and communities

Reddit's role seems to be opinion, not discovery. In one August 2026 capture, ChatGPT searched Reddit about six brands it had already picked. Reddit threads made up 84 of the 221 pages in that pool, and none were credited in the answer. So threads may shape how ChatGPT describes you more than whether it names you.

Show up where buyers in your category ask for advice, answer the real question, and say who you are. Rankbox's Reddit Presence, rolling out now, finds relevant threads and drafts replies for you to post yourself.

YouTube and wider brand mentions

YouTube mentions had the strongest link to AI visibility in Ahrefs' 75,000-brand study, at about 0.737 across ChatGPT, AI Mode and AI Overviews. That's a correlation, not proof of cause. But reviews, demos and interviews on YouTube are also the kind of coverage that feeds the slow loop. Our glossary entry on brand mentions explains why a mention counts even without a link.

Make Your Own Pages Answer the Sub-Queries

Your own site does its work after ChatGPT has named you. Its vendor checks go straight to your domain for official facts. If the page isn't there, or is vague, the check comes back empty. So a clear site is still part of how you rank on ChatGPT.

Four rules make these pages work:

1. One strong page per intent. In Mohanadasan's data, two pages from one domain in a result group had the best citation rate, at 6.2%. Six or more fell to 1.7%. Merge near-duplicates.
2. Keep your domain obvious. ChatGPT fills in the domain for its vendor check from memory. In one capture, it probed a domain the brand doesn't run. After a rebrand or migration, keep redirects and your name consistent everywhere.
3. Keep buying pages current. In those captures, vendor checks on buying questions carried the number 30 in a field the author reads as a recency window in days. If he's right, a pricing page untouched for a month may miss out.
4. Let the crawler in. ChatGPT can only read what OpenAI's search crawler can reach and what sits in the HTML. Our ChatGPT SEO guide covers the settings, and the free AI search readiness check scans a URL for the basics.

How you write each page is its own craft. Our guide to how to optimize content for AI search covers passage-level writing in depth.

Product Carousels and Ads

Product results for physical goods

If you sell physical products, there's a second way to rank on ChatGPT: the product carousel. OpenAI says product results are "selected independently by ChatGPT and are not ads." They draw on structured metadata from first-party and third-party providers, plus other third-party content. When a buyer opens a product, merchants are ranked on availability, price, quality and whether they're the maker or primary seller.

Getting into that data is a separate job from your content. As of September 2026, Shopify merchants are already included through Shopify Catalog. Others can apply to send a direct product feed. OpenAI has also expanded its Agentic Commerce Protocol (ACP, the standard merchants use to share feeds) to product discovery, and now lets merchants use their own checkout while it focuses on discovery.

Ads don't buy a place in the list

OpenAI began testing ads in ChatGPT in the US on 9 February 2026, for Free and Go users only. Ads sit below the answer, carry a sponsored label, and OpenAI says they "do not influence ChatGPT's answers." Paid placement is a separate channel with its own budget, not a shortcut into the list.

How to Check Where You Rank on ChatGPT

You can see where you rank on ChatGPT today in an afternoon, with no tool:

1. Write 10 to 20 unbranded prompts. Use the shapes buyers use: best X for Y, alternatives to a rival, X for a use case. Our free AI Visibility Prompt Kit drafts 30 buyer prompts and a scorecard.
2. Run each prompt five to ten times. Use Temporary Chat, which OpenAI says doesn't use or update memory, and stay on the same plan and model.
3. Record each list in order. Note every brand and its position, not just whether you appear.
4. Score it. Work out mention rate, average position and top-3 rate per prompt, as in the Plannora example.
5. Read the first query if you can. Mohanadasan showed how to find it in your browser's developer tools. OpenAI changed the format in August 2026, so expect it to move again.
6. Repeat monthly. Keep the prompts fixed so changes mean something.

For sample sizes and a control test that separates your work from model drift, see our guide on how to measure GEO. Rankbox doesn't track AI citations today, so this manual panel, or a paid tracker, is how you'll see whether your rank moves.

How Long It Takes to Rank on ChatGPT

Two loops mean two clocks for anyone trying to rank on ChatGPT.

Expect the fast loop to flicker. In Ahrefs' experiment, the self-promotional pages that got cited were cited on only about one in three eligible days between their first and last appearance. Measure over weeks, not days.

Expect the slow loop to favor whoever is ahead. In Semrush's study, clear category owners stayed on top in 90.4% of month-over-month comparisons. Where the leader flipped, the median lead was just 1.3 percentage points. So if your category is still unsettled, now is the cheapest time to rank on ChatGPT for it.

Where Rankbox Fits in the Shortlist Loop

Rankbox helps you rank on ChatGPT through the parts of the loop you control from your own site. Answer-Space Research maps the questions buyers ask ChatGPT, Perplexity and Google, and scores them for estimated volume, difficulty and intent, so you know which "best X for Y" prompts to target. The Citation-Ready Writer then researches the live web and writes source-backed articles, like the comparison and use-case pages above. They reach your site through Rankbox's API, which a developer wires in.

Rankbox doesn't track AI citations or rankings today, so pair it with the free Prompt Kit or a tracker to see whether your mention rate moves. One plan covers it: Business, at $49.50 a month for one site and 30 articles, with a 7-day trial. See pricing.