Clients are asking why they don't appear in ChatGPT answers: where agencies should start
Clients ask your agency why they don't appear in ChatGPT answers and you have no process? Here is the repeatable workflow we run, from prompts and access to citations.
You will recognize the email. A client tried a prompt, did not see their brand, and wants to know why. If your agency has no standard answer, each of these becomes a one-off scramble that eats a day and ends in a vague reply.
We built a routine for exactly this, and it now runs the same way for every client. Here is where we start, in order.
Start by turning the screenshot into a prompt set
One screenshot is one prompt, asked once, in one place. Before diagnosing anything we agree a set of 25 to 50 prompts with the client, in their customers' language. We cover four groups: category questions, comparison questions naming competitors, problem questions and brand questions.
We tag each by topic and, where it matters, by persona and country. Different personas ask differently, and answers differ by location. Prompt Volumes and difficulty scores help us choose which prompts deserve tracking, and the query fan-outs show what the engine searched for before it answered.
Do not skip this. It changes the client conversation from "ChatGPT ignored us" to "we appear in this share of the prompts that matter, and here is where we do not."
Take a baseline before you change anything
We run the set daily across the engines the client cares about and record mention, position in the answer, sentiment, the competitors named and the sources cited. Answers vary between runs, so one reading proves little. We look at the trend over a couple of weeks.
Check access before content
This is where most agencies waste months. They commission articles for a site that the relevant bots cannot fetch.
We read the crawler logs. Which AI bots visit, which pages, what response codes? OpenAI documents several crawlers with different purposes, so we read its current documentation before touching robots.txt, then confirm in the logs that the change took effect. We also check the CDN, firewall and bot-protection rules. Well-meant security settings are a frequent cause of blocked crawlers.
Read the citations like a map
For prompts where the client is missing, we list the pages and domains the engine cited instead and group them: competitor pages, review sites, editorial, forums, video. The mix tells us what to do.
- Review sites dominate: the client needs presence and reviews there.
- Forum threads dominate: the client needs to take part in the discussion honestly.
- Competitor comparison pages dominate: the client needs its own comparison content.
- Editorial dominates: the work is digital PR.
Offsite mentions, where a brand is named on a third-party page without a link, matter here too, and they are easy to overlook.
Fix in a fixed order
Access, then content gaps, then offsite presence. We keep a short visible list so the client sees movement. Typical items: unblock a crawler, publish a page that answers a buyer question directly in plain text, fix errors on key pages, get listed where the engines already look.
Measure outcomes, not only mentions
Mentions are interim. We attach the client's analytics so AI-referred visits and conversions show in the same report. That is what turns an AI visibility retainer into something a finance director understands.
What we run it on
We run this on Promptwatch, and the match to the workflow is the reason:
- Prompt set: Prompt Explorer, Prompt Volumes, personas, topics and tags.
- Baseline: Visibility Score, Share of Voice, Sentiment and Position in Answer with daily monitors.
- Access: Agent Analytics, which gives IP-verified crawler logs and error tracking with explanations.
- Citations: citation analytics at page and domain level, offsite mentions, plus Reddit, YouTube and LinkedIn reports.
- Fixes: the Action Board for a prioritized to-do list, content gap analysis, and publishing to WordPress, Webflow or Framer.
- Outcomes: visitor analytics and the 360 Insights view that joins crawls, citation clicks and visits per page, with scheduled PDF reports.
On agency plans (Kick-off $199/mo, Growth $399/mo, Scale $799/mo) we get unlimited projects and prompts, 10 seats, crawler logs and custom reports. They do not include the automated Content Agent or AEO articles, so we plan content production around that.
The client email
Once the first week's work is done, the answer to the original question is short. Here is what we found. Here are the two things we will fix first. Here is when you will see the next report.
If the client pushes for a guarantee, we say it plainly: no agency and no tool can guarantee that ChatGPT will recommend a brand. We can find and remove what is in the way, and show the effect.
Make it a service, not a scramble
Write the routine down, price it, and give it a name. The first audit takes the most time. Once the prompt set, baseline and access check are templated, they take far less. If you want to see our version of the audit, ask us and we will walk through a sample.