ChatGPT6 min read
ByKarol PiaseckiFounder TextRank.ai
Published
Reading time6 min read

ChatGPT Not Recommending Your Company? How to Change It in 3 Steps

ChatGPT recommends competitors for a buyer question? Find out why, get a page-level plan and see how the cost compares with agencies and monitoring.

A prospect on a demo says: "I asked ChatGPT for a tool that does what you do. It recommended three companies. You weren't one of them." You type the same question and see the same three names. Your site exists, it has content, it even gets Google traffic. For this question, the model does not choose it.

You can find out why. You need three things: the exact question, a comparison of what ChatGPT cites against what your page says, and a plan someone will implement. Below is each step, followed by detailed costs and how they compare with the alternatives.

Step 1: Write down the prompt you are losing

Without a specific question, there is nothing to fix. "ChatGPT doesn't recommend us" is a feeling. "For 'fleet management software for a 30-vehicle company with fuel card integration', ChatGPT recommends three competitors and cites an industry comparison site" is a problem you can work on.

Where to get the question:

  • sales conversations: ask new customers what they typed into ChatGPT before finding you, and write down their words, not yours,
  • Google Search Console: long queries (six words or more) show how buyers describe their situation,
  • Bing Webmaster Tools: the AI Performance report lists the queries the model used to retrieve sources (Bing).

We describe how to collect this data step by step in how to build an AI search audit in 5 steps.

Before moving on, run two quick checks.

Ask the question yourself. Use ChatGPT with search on, in a new chat with no history about your company. Record the date, the full answer, the brands named and the source links. Ask two variants of the same question too. If the same competitors win all three, you have a pattern, not a fluke.

Check robots.txt. OpenAI uses a separate crawler, OAI-SearchBot, for ChatGPT search. If it is blocked, your pages will not appear in search answers regardless of content quality (OpenAI documentation). Blocking GPTBot, which collects training data, does not affect search. Changes take effect after about 24 hours.

Step 2: Put the prompt into TextRank and see what your page is missing

By hand, you would open every source from ChatGPT's answer next to your page and list what they say and you do not. With five sources that is an hour or two per question, and the result depends on who reads.

In TextRank you enter the prompt, pick a model and give the URL of the page that should answer the question. The audit:

  1. asks the model and records the answer and the sources it cites or mentions,
  2. checks whether the model can fetch your page at all,
  3. compares the sources with your page: which entities, claims and passages they have and you do not,
  4. builds a plan: each gap gets a move (keep, rewrite, add an answer, get proof, create supporting material or park the prompt), a priority and a reason.

Say you sell fleet management software. For your question, ChatGPT cites a competitor's pricing page and a ranking on an industry site. The audit shows the competitor's pricing page lists, in a table, the vehicle count where each plan starts and which fuel cards integrate. Your page says "integrations with popular fuel cards" and gives no threshold. The recommendation: add a table of thresholds and supported cards to your pricing page, once the product owner confirms them. The industry ranking becomes a separate task: you cannot fix it on your own site.

That distinction matters. Some causes sit on your pages, some in places you have limited influence over. The plan separates them so your team does not rewrite a page when the real gap is missing reviews.

Step 3: Implement the plan and check the same question again

You export the plan to Markdown or PDF. Each recommendation points to the prompt, the model's answer, the source and the exact passage on your page, so it can be implemented by:

  • you, if you edit the site,
  • a copywriter or agency, who gets a ready brief instead of a vague "let's work on AI",
  • an AI tool (Claude, ChatGPT, an agent in your site repository) given the Markdown file as context. Ask it to draft one change at a time and do not let it invent product facts.

Facts are always approved by someone who knows the offer. The audit will show that a pricing threshold is missing; it will not decide it for you.

After publishing, run the same audit for the same prompt and model. Compare sources, not only whether your name appears. ChatGPT answers vary between runs, so one mention proves little. You are looking for a change that holds across checks.

What it costs

One audit is one question checked in one model. A $25 starter pack with no subscription covers roughly 4–7 audits, depending on the model. Premium at $99 per month covers roughly 20–35 audits a month. See current pricing.

One pack is enough to start: check the question from the demo, make your changes, run the retest and you still have room for a few more questions. If you want to check a dozen or more questions in two models on a regular basis, Premium is the easier option.

Add the time of whoever implements the changes. For one page and one recommendation that is usually an hour plus fact approval, but it depends on the page and your process.

How it compares with the alternatives

ApproachWhat you pay forStarting cost
By handTime: answers, sources, comparison, brief$0 plus several hours per question
AI monitoring: Otterly.AIDaily tracking of a prompt setfrom $29/month for 15 prompts; Claude as an add-on from $29/month (pricing)
AI monitoring: Peec AIPrompt tracking, visibility, sourcesabout $95/month for 50 prompts according to third-party listings (Rework)
AI monitoring: ProfoundPlatform for large brandsfree trial, Enterprise plan priced on request (pricing)
AgencyAnalysis, editing, re-measurementa small project typically runs into four figures (cost breakdown)
TextRankOne-prompt audit with a page-level plan$25 pack, no subscription

We checked competitor prices on 19 October 2026. Peec AI does not show amounts on its public pricing page, so we cite a third-party listing.

Monitoring tools and TextRank solve different problems. Monitoring answers "how often do we appear across hundreds of prompts, day after day". If that is what you need, the per-prompt price in a monthly plan can be lower than a single audit. TextRank answers "what exactly should change on this page to have a chance of winning this question". Once you know which question you are losing, you do not need a monthly tracking plan to get a fix: you pay per audit with no subscription. That is less than the cheapest monitoring plan and far less than an agency, as long as someone on your team implements the changes.

For a wider comparison of approaches, read AI visibility monitoring vs content audit vs SEO and EchoRanked vs Peec AI vs TextRank.

When three steps are not enough

If ChatGPT cites only rankings, reviews and forums for your question, your page alone will not be enough. You need presence in those places, and the audit will only show you which ones. If your product genuinely does not meet the condition in the question, no content change will hide that, and none should. Pick a question where you have an honest advantage.

Start with one question

Take the sentence you heard from a prospect, check your robots.txt and run an audit in TextRank. To see what you get, read the example audit from prompt to page change.