ChatGPT SEO: How to Track and Rank Your Brand in ChatGPT

Most advice about ChatGPT SEO jumps straight to tactics: add schema, write FAQs, get onto a listicle. Skip a step and you optimize blind. You want your brand to come up when someone asks ChatGPT for a recommendation in your category, and that sounds like one goal, but “come up” has two very different meanings. ChatGPT can name your brand in the answer with no link, or it can cite your page as a clickable source. Across about 4,800 ChatGPT answers we logged for a sample of brands in competitive categories, of the answers where ChatGPT brought a brand up at all, roughly a third only named it, with no source attached. So before you change a single page, the first job is to find out which kind of presence you actually have.

Measure first: the three signals worth tracking

You cannot improve what you have not measured, and on ChatGPT three signals carry almost all the meaning.

• AI Mentions: does ChatGPT bring your brand up at all, whether by naming it or citing it.
• Brand Coverage: across the prompts you care about, the share of answers where you appear. It is a rate, not a count, so a brand tracked on 100 prompts and present in 60 is doing better than one present in 80 of 200.
• Cited or only named, and who shows up instead: whether you are a clickable source or just a name in the prose, and which competitors and which URLs ChatGPT cites when it answers your prompts.

That third signal is the one people miss. The split in our data surprised us. We expected ChatGPT to behave like Perplexity, which almost always attaches a source when it surfaces a brand. It does not. On ChatGPT, about a third of the answers that mention a brand never cite it, and that share held up across most of the brands in the sample, so it is not one odd account dragging the average.

The difference changes what you do next. A name in the prose with no link usually comes from the model’s trained memory, which lags the live web by months and sends you no traffic. A citation means ChatGPT’s search system fetched a real page and chose yours. If you are getting named but not cited, your fix is to become a page the search step can reach and quote, not to wait for the model to remember you better.

How to track your brand on ChatGPT by hand

For a handful of prompts you do not need any tooling. You need a method you run the same way every time.

1. Write down 10 to 20 prompts a real buyer would type. The ones that trigger a web search lean commercial: “best [category] for [use case]”, “[your brand] vs [competitor]”, “[category] alternatives”. Informational questions often get answered from memory alone.
2. Open ChatGPT with web search on and ask each prompt in a fresh chat, so one answer does not prime the next.
3. For each answer, record three things: did it name your brand, did it cite one of your pages as a source, and which competitors and URLs it cited instead.
4. Ask each prompt two or three times across different days. Answers shift from run to run, so a single check is noise, not a reading. If a result swings wildly, that is the model’s normal variance, covered in our visibility-drop diagnostic.
5. Tally Brand Coverage as the prompts where you appeared over the prompts you tested. That single rate is your baseline.

This stops scaling at around ten prompts. Three runs each, repeated weekly, across more than one assistant, and the manual sheet eats an afternoon you do not have, while still showing you a snapshot instead of a trend line. That is the point where automated tracking earns its place: it watches the same prompts on a schedule and surfaces the competitor and source detail the manual pass cannot hold.

How to rank on ChatGPT

Ranking on ChatGPT is not a single setting you flip. It is a loop: set a baseline, pick your targets, make sure you are reachable, study what wins now, rebuild your page to beat it, then re-measure.

1. Set the baseline. Use the numbers from the tracking pass above. Without them you cannot tell whether anything you change later actually helped.
2. Pick ten prompts to own this quarter. Coverage on every prompt is a fantasy. Choose the ten where you have a genuine claim to be the answer, and ignore the rest for now.
3. Confirm ChatGPT can reach your content. This is the cheapest fix and the most commonly broken one. OpenAI runs three separate crawlers, and they are not interchangeable.

The one that decides whether you can be cited is OAI-SearchBot, and plenty of sites block it by accident while trying to keep their content out of AI training, per OpenAI’s own crawler documentation. Check what you are actually allowing with our robots.txt AI bot checker, and read why crawl access comes before any other AI work if this is new to you.

Reachability is only half of it. ChatGPT search also draws on the Bing index, so once OAI-SearchBot can crawl you, the fastest way to push a new or updated page toward ChatGPT is IndexNow, a one-ping protocol set up through Bing Webmaster Tools that notifies Bing in minutes instead of waiting on a crawl. That lever is specific to the Bing family; each assistant reads a different index, which we map in which index each AI assistant reads.

4. Find the URLs ChatGPT cites for your targets. Your tracking already logged them. Those pages are the bar you have to clear, so read them as the standard the answer is currently being held to.
5. Rebuild your page to beat them. Lead with a clear, quotable answer near the top, carry the specifics the current sources have, and structure it so a machine can lift a clean claim out of it. Keep in mind that AI usually describes a brand from third-party and competitor pages, not its homepage, which we measured in when AI describes your brand, it cites your competitors; sanity-check a draft with our content readiness checker.
6. Publish, then wait a full check cycle before judging. A Daily Check shows you recent movement; a Full Check is the baseline you compare rank against. Do not react to a single answer the day after you ship.
7. Scale what worked, drop what did not. When a prompt moves, apply the same rebuild to the next ten. When it does not, the page was not the bottleneck and you go back to the reachability and competitor checks.

Three mistakes that waste the effort

Most stalled ChatGPT optimization comes back to one of these.

• Stuffing FAQ schema and hoping. Structured data helps a machine parse your page. It does not make a thin answer worth quoting, and ChatGPT will still pass over a page it cannot extract a real claim from.
• Publishing AI-generated filler to “feed the model.” ChatGPT’s search step retrieves many pages and cites only the few it can attribute a specific, trustworthy claim to. Generic copy gets fetched and discarded, so volume for its own sake is wasted budget.
• Treating ChatGPT as the whole job. The assistants barely cite the same sources, as we found when we compared citations across five engines. Winning ChatGPT and ignoring Perplexity, Gemini, Claude, and Google AI Mode leaves most of your AI visibility unclaimed.

Where to start this week

Run the tracking pass on ten prompts first, so every later change has a baseline to beat. Before you touch any content, confirm ChatGPT’s search crawler can reach the pages you want cited with our robots.txt checker; a surprising share of “why is ChatGPT ignoring us” problems end right there. Once access is clean, look at the third-party and competitor pages AI already cites about you, the pattern in when AI cites your competitors, and pick up the broader loop across every assistant in our AI visibility playbook.