How can I see if ChatGPT recommends my company? A practitioner's method

Almost every new client asks us to "check ChatGPT" for their company. We understand why. It is quick, a little nerve-wracking, and the answer feels decisive. It is not decisive, and the first thing we do is explain why before we run anything.

One answer is a sample

ChatGPT answers change with phrasing, location, conversation history and the day you ask. Whether search is on matters too: with search, it browses and cites sources, and without it, it answers from model memory. So we never present one result as a verdict. We present a set of prompts, run repeatedly, as a range.

The method we start with

Write real buyer prompts. Pull them from sales calls, support tickets and the questions prospects ask in discovery. Group them: category, comparison and brand. If every prompt includes the company name, the test is meaningless because the model only has to confirm you exist.

Run each prompt in a clean chat. Note the mode. Record whether the brand appears, whether it is recommended or merely listed, its position, which competitors appear and which sources are cited.

Read the sources. This is the most valuable step. When ChatGPT leans on a review site, a forum thread or a competitor's comparison page, you learn where the decision is made. Often the fix is not on your own site.

Check access. If OAI-SearchBot is blocked, ChatGPT search cannot fetch your pages. GPTBot and ChatGPT-User serve different purposes and are governed by separate rules, so we read the robots.txt and firewall config before drawing conclusions.

When we move off the spreadsheet

Manual testing breaks around fifty runs. It also cannot see crawler hits or AI referrals. At that point we use Promptwatch for client work.

The features that matter here: prompt panels run daily per model, country and language, so we get a trend, not a snapshot. Visibility Score, Share of Voice, Sentiment and position in the answer give us numbers to compare month over month. Citation analytics shows which pages and domains ChatGPT cites. Agent Analytics shows IP-verified crawler logs for GPTBot, OAI-SearchBot and ChatGPT-User, with failure explanations, which settles the access question with data. Visitor analytics connects ChatGPT referrals to conversions.

Two free aids are worth knowing. The GEO Inspector Chrome extension shows the fan-outs ChatGPT ran and which sources it fetched versus cited, and runs locally. The AI Visibility Report scans without a login. Both are useful for a first look.

Plan notes

The free Explore plan is ChatGPT only with 10 prompts and 310 responses, a good fit for exactly this question. Essential at $95 a month adds more prompts but no crawler logs. Crawler logs start on Professional at $245. Agency plans start at $199 and suit people running this for several companies.

What we tell clients about "recommended"

Being mentioned is the low bar. A recommendation reads as a reason: "good for teams of five to fifty because..." We look for that language and for which attributes ChatGPT ties to the brand. If the attributes are wrong or outdated, the fix is often a clearer page, a corrected profile on a source ChatGPT cites, or more third-party coverage.

We also say plainly that no tool or agency can make ChatGPT recommend a company. What we can do is find the cause of a gap, fix what is fixable and measure again.

A first-week checklist

Write ten prompts across the three groups.
Run them twice, a few days apart, and record both.
List the sources cited and flag any you could influence.
Check robots.txt for OAI-SearchBot, GPTBot and ChatGPT-User.
Load the prompts into a free Promptwatch Explore account to keep a daily record.

By the end of the week you will know whether you have a visibility problem, an access problem or a content problem, and each one has a different fix.