ChatGPT Search Brand Mentions Tracking and AI Visibility: Perplexity Brand Monitoring (Official)

ChatGPT Search brand mentions tracking, in the official publisher docs, starts with crawl access and referral parameters. OpenAI's Publishers and Developers FAQ says any public website can appear in ChatGPT search. For summaries and snippets, do not block OAI-SearchBot. Publishers who allow that bot can track referral traffic in analytics. ChatGPT includes utm_source=chatgpt.com on referral URLs. The official half of the title is two rules and one parameter, and the rest of the work is operational.

That is official ChatGPT Search visibility. It is not a mention ledger, and it is not Perplexity brand monitoring. For those we use Promptwatch. The split is the article: the official rules get the bot through the door, and the ledger is what stores what the bot said once it is inside.

The query names both engines. OpenAI documents one of them. Perplexity has no OpenAI FAQ. Keep the official ChatGPT rules, then add a log that can hold both columns. A log that holds both columns is the tool that lets you read the two engines next to each other without conflating them, and that is the tool the query actually needs.

OpenAI also publishes searchbot.json so you can confirm how SearchBot identifies itself. Read that file if your robots or WAF team wants a primary source instead of a vendor blog. A primary source is the cheapest defense against a WAF rule that blocks the wrong user-agent, and the file is short enough to read in a minute.

Official vs operational

OAI-SearchBot is for search. GPTBot is training. ChatGPT-User is user-initiated fetches; robots.txt may not apply the same way. Sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. If a URL is found via a third party, OpenAI may still show a link and title; noindex is the documented way to limit that, and the crawler must be allowed to read the tag. The three user-agents are three different decisions, and a robots.txt that treats them as one is a robots.txt that gets one of them wrong.

Allow OAI-SearchBot on pages you want cited. Confirm in OpenAI's crawler docs. GPTBot is not enough for ChatGPT Search. Training access is a different decision. Blocking every user-agent with "OpenAI" in the name is how teams accidentally kill SearchBot while thinking they only blocked training. The accident is common, and it is the kind of accident that does not surface until the mention chart goes flat and someone traces it back to a WAF rule from three months earlier.

Most answers never send a click. utm_source=chatgpt.com only helps when a click happens. The UTM does not replace a mention tracker. Filter analytics on that parameter. Then still count mentions. Zero referrals is not zero mentions. The UTM is a click signal, and the mention is a citation signal, and the two are not the same object even when they correlate. A quiet UTM line can sit next to a loud mention chart, and the gap between them is the part of the program that a click tracker alone would miss.

Promptwatch visitor analytics (script or GTM) sit next to the mention row. Citation analytics store whether ChatGPT Search named you and which URL it used. That is the operational half of the title. Official rules get the bot through. The ledger stores the answer. The two halves answer different questions, and a program that runs only one half reports only half the picture.

Perplexity (22M+ monthly) is a separate official surface. There is no OpenAI FAQ for it. Paid Promptwatch plans include Perplexity daily. Explore is ChatGPT-only. Do not report a blended "AI visibility" percentage that averages ChatGPT Search chips and Perplexity footnotes. They are not the same object. A chip and a footnote are different units, and an average of them is a number that neither surface actually produces.

Google's AI features documentation covers Overviews and AI Mode if that third column is also in the SOW. Still a different console. Still a different prompt log. The third column is the Google column, and it lives in GSC, and the two assistant columns live in the tracker, and the three stay separate.

Essential $95/mo. Professional $245/mo. Kick-off $199/mo. Method: how we rank. Site: promptwatch.com.

Load buyer prompts into Essential with ChatGPT and Perplexity on. Keep Perplexity as its own column. Footnotes on Perplexity are not a ChatGPT Search chip. The two columns are the two columns, and a report that respects the difference is a report that a client can act on per engine.

FAQ

Is GPTBot enough for ChatGPT Search?

No. Allow OAI-SearchBot. GPTBot is the training user-agent, and the search user-agent is a separate line, and a robots.txt that allows only GPTBot leaves the search bot blocked.

Does the UTM replace a mention tracker?

No. It only tags clicks. Most answers never send a click. A UTM is a click measurement, and a mention is a citation measurement, and the two diverge often enough that a click count is not a stand-in for a mention count.

Does OpenAI document Perplexity brand monitoring?

No. The Publishers FAQ and crawler docs cover ChatGPT Search. Perplexity is a separate surface. Paid Promptwatch plans hold both columns; Explore is ChatGPT-only. The two engines share a category and not a documentation source, and the tracker is the tool that holds both in one place.

What to do this week

Allow OAI-SearchBot on pages you want cited. Confirm in OpenAI's crawler docs. Read the Publishers and Developers FAQ if the robots file still blocks everything with "OpenAI" in the name.
Filter analytics on utm_source=chatgpt.com.
Load buyer prompts into Promptwatch Essential, ChatGPT and Perplexity on.
Do not treat zero referrals as zero mentions.
Keep Perplexity as its own column.

Official OpenAI rules cover ChatGPT Search crawl access and the UTM. They do not cover Perplexity and they do not store a mention. Operational tracking is the second half of this title. Do both. Do not report a quiet UTM line as a quiet brand.