Most teams that want clients from ChatGPT, Gemini or Claude start by asking what an AI search audit costs for their whole website. That is the wrong unit. A buyer does not evaluate your domain. They ask a question, read an answer and decide which vendors to contact.
Budgeting per buyer question keeps money on the answers closest to revenue. This guide shows how to price 5, 10 or 20 questions, what to do with them in the first 30 days and how to tell whether AI answers are bringing clients.
Which questions bring clients
Not every prompt is worth paying for. The ones that bring clients share one feature: the buyer is about to act. In practice, they fall into six intents.
| Intent | Example buyer question |
|---|---|
| Choice | Which helpdesk tool fits a 10-person B2B support team? |
| Price | How much does a helpdesk with SLA reporting cost per agent? |
| Implementation | How long does it take to migrate from Zendesk? |
| Comparison | Freshdesk vs your product for SaaS support |
| Integration | Which helpdesk integrates with Linear and Slack? |
| Risk | Can I export all tickets if I cancel? |
Informational questions such as “what is a helpdesk” can build awareness, but they rarely change which vendors land on a shortlist. Start with the six intents above. Read how to choose buyer prompts for an AI audit for the selection method.
AI search audit pricing per question: 5, 10 or 20 prompts
TextRank prices one audit as one buyer 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.
| Scope | ChatGPT only | All four models |
|---|---|---|
| 5 questions | one starter pack | fits in Premium |
| 10 questions | two starter packs | Premium plus a few extra packs |
| 20 questions | fits in Premium | spread over several months |
Two rules keep this budget efficient:
- Go wide in one model first, deep in four models later. Check all questions in the model your buyers use most. Run all four models only for the questions where the first result shows a real gap.
- Reserve budget for the retest. Every question you change should be checked again. Plan about a third of the budget for the second round.
The audit fee is usually the smallest line. Writing, fact-checking and publishing cost more. For a full annual view, see the AI visibility audits cost guide.
A 30-day plan from question to client
Days 1–3: choose five questions. Pull them from late-stage sales calls and support tickets. Assign each one to the page that should answer it: pricing, a product page, documentation or a comparison page.
Days 4–7: run the first audits. Check all five questions in one model. Read the sources the model used before you open your CMS. Note which questions show a gap you can close on your own site.
Days 8–10: deepen the top two. For the two questions closest to a purchase, run the remaining models. You will see whether the gap is consistent or specific to one model.
Days 11–21: ship three changes. Take the highest-priority recommendations. In TextRank, each one comes with a move — rewrite, add an answer, get evidence, prepare a supporting asset — plus the passage it concerns. Have the product owner confirm every fact before publishing.
Days 22–30: retest and record. Rerun the changed questions in the same model. Compare sources, not just mentions. Keep the old and new reports side by side for the next round.
For five questions, the plan can be run by one owner alongside other work. If no one can write and publish, budget for that capacity before you buy audits.
How to measure whether AI brings clients
AI-sourced demand is hard to attribute because the buyer often reads the answer, closes the chat and types your URL later. Use several signals together:
- Self-reported source. Add “Where did you first hear about us?” to demo forms, with ChatGPT, Gemini and Claude as options.
- Referral traffic. Some visits from AI assistants arrive with a referrer or a tag such as utm_source=chatgpt.com. Many do not and land in Direct, so treat this number as a floor, not a total.
- Answer and source changes. Retested audits show whether your page has started appearing as a source for the questions you changed.
- Sales conversations. If prospects stop asking questions your page now answers, the page is doing its job before the call.
None of these is proof on its own. Together, they show whether the questions you invested in are moving buyers toward you.
What an audit will not do
An audit shows why a model chooses other sources and what to change on your pages. It does not publish content, earn third-party reviews or guarantee that a model will cite you. If the winning sources are all independent review sites, the right next step may be review outreach rather than another page edit.
Start with five questions
Write down the five questions your best customers asked before they bought. Run them in one model and fix the first gap.
Price your first five buyer questions in TextRank. For the full method, follow the five steps to be visible in AI search, and see how this compares to paid channels in AI lead generation vs organic AI search.