Prospects tell our sales team they found our competitors through ChatGPT, but we never come up: how to fix that
Prospects find your competitors through ChatGPT and you never come up. Our agency diagnostic: reproduce the prompt, read the sources, check bot access, fill gaps.
This is one of the most common first conversations we have with a new client. Their sales team keeps hearing prospects say they found a competitor through ChatGPT, and the client has never once appeared in those answers. There is usually some panic in the room. We try to turn it into a checklist.
The order matters, because the cheapest explanations come first.
Reproduce the moment
We ask the sales team for the real wording prospects use, or the nearest thing. The patterns are familiar. A category question with a company type attached. A request for alternatives to a named competitor. A question about which tools work with a specific system.
We run those in ChatGPT with search enabled and write down who is named, in what order, and what the answer cites. Then we turn the questions into a tracked prompt set, split by topic, persona and intent, because a single run is only a sample. Answers change from day to day.
We run client programs on Promptwatch for this stage. Its prompt panels run daily per model, country and language, and persona stacking lets us ask the same question as a buyer in different roles. The competitor heatmap shows quickly where rivals are present and the client is not.
Read the sources, not just the answer
A named competitor is the result of pages somewhere making their case. So the next question is which pages. Citation analytics show the cited pages and domains for each prompt. In our B2B work we see three recurring patterns. Rivals appear in review and comparison pages the client has never been listed on. Rivals have community threads, often on Reddit, where they are discussed in the buyer's own words. Or the engine cites the rival's own help-center pages, which answer the question more directly than the client's.
Promptwatch's offsite mentions view helps with the first pattern, since it shows where a brand is named on third-party pages without a link. The Reddit, YouTube and LinkedIn citation reports cover the second.
Rule out a crawler problem
Before we write a word of content, we check whether the bots can reach the site. We have seen programs where the first useful finding was access, not content. Agent Analytics logs real-time, IP-verified AI crawler hits, so spoofed user agents are filtered out. We look for GPTBot, OAI-SearchBot and ChatGPT-User, and for errors with failure explanations.
Promptwatch publishes a case study from its customer TreeHouse, the company behind Pararius and Huurwoningen.nl. AI search grew to about 25% of its organic conversions, 99% of it from ChatGPT, after it fixed crawler access and automated weekly reporting through the Promptwatch MCP. That belongs to that customer and we do not present it as something every client will get. We do use it to explain why access comes first.
Crawler logs begin on the Professional plan at $245 a month, so a client on the entry plan will not see them.
Fill the gaps that are real
If access is fine and sources show a gap, we move to content. Content gap analysis compares what the engine looks for with what the client has published. Then there are three choices, which we weigh case by case: refresh an existing page, publish a new one, or earn a mention elsewhere. Content Refresh produces diff-style suggestions for existing pages, which is faster than rewriting from scratch. For new pages the Content Agent on Business and Enterprise can plan, write and publish against gaps, with a review inbox so a human approves each piece.
We also look at Search Console and Bing data. Official guidance for AI features in both searches points back to crawlable, helpful pages with sensible structure, so a page that has trouble in classic search often has trouble here too. Search Insights in Promptwatch overlays that data on AI crawls and citations.
Close the loop with the sales team
We ask sales to log the "how did you hear about us" answer in the CRM verbatim. It is a useful check next to the tool data, and it is how we know whether a change is reaching real buyers. Visitor analytics adds the measured side, attributing AI-referred sessions and conversions through a script, GTM template or WordPress plugin.
What we tell clients up front
We do not promise a date or a ranking. Nobody can guarantee that ChatGPT will name a brand, and the engines change without notice. We promise a diagnosis in the first weeks, a ranked list of actions, and a monthly read on whether the prompts we care about are moving.
Our default for this problem is simple. Run the prompts, read the citations, check the logs, then write. Most pure trackers cover only the first step. The crawler logs and traffic attribution are the part we would not want to work without.