A competitor appearing in one ChatGPT answer is a reason to investigate, not proof that they dominate AI search. Answers vary with the wording, platform, date and user context. A useful competitor audit uses a repeatable question set, records sources and connects findings to pages you can improve. It should never turn a handful of prompts into an invented market share figure.
Choose competitors and questions carefully
Start with the businesses a customer genuinely compares with you. Include direct local competitors and, where relevant, publishers or directories that often answer the same questions. Do not choose rivals only because an AI tool mentioned them once.
Build the question set from customer enquiries, sales objections, site search and commercial search queries. Include:
- Basic information questions that establish the category.
- Comparison questions where a buyer weighs options.
- Local questions with a Hong Kong district or service constraint.
- Questions about price factors, process and suitability.
- Follow up questions that reveal whether the answer stays consistent.
Keep the wording and date in the audit. A tiny change can change the answer, so comparisons require a consistent method.
Record what actually appears
For each test, capture the platform, exact question, answer date, brands named and URLs cited. Note whether a brand appears as a recommendation, a neutral example or a warning. Record when no sources are shown. Avoid treating a linked article as an endorsement of the company that published it.
| Observation | Useful next question | Possible action |
|---|---|---|
| Competitor cited for a practical guide | What detail makes that guide useful? | Improve your own relevant guide |
| Directory cited repeatedly | Is your business information accurate there? | Correct or complete a legitimate listing |
| Your page cited but no enquiries | Does the page offer a relevant next step? | Improve navigation and conversion path |
| Inconsistent answers | Does wording or platform change the result? | Expand the sample before making a decision |
Our AI search optimisation guide explains why page quality and discoverability remain important.
Audit the cited pages, not just the brand names
Open the cited URLs and look for the answer’s evidence. Does the page explain a process clearly? Is it dated? Does it include original examples, useful comparisons or relevant qualifications? Is the author or business identity clear? Look for strengths you can adapt to your own expertise, not text to copy.
Then inspect the page on your site that should answer the same question. It may be missing, hard to find, thin, outdated or aimed at a different intent. Our search intent guide helps diagnose that last case. Sometimes the right response is to improve an existing service page rather than publish another article.
Check public business information
If an assistant describes a competitor more clearly than it describes you, inspect your About page, service descriptions and public listings. State what the business actually does and where it operates. Do not manufacture awards, reviews or claims merely to fill a perceived gap.
Turn findings into a short work plan
Sort opportunities by customer value and confidence. Correct factual errors first. Next address pages that support high value enquiries. Add missing examples, clearer comparisons and relevant internal links. Test technical access if important pages never appear. Assign an owner and a review date to each action.
Repeat the answer checks later using the same questions, but do not claim a page edit caused an answer change without stronger evidence. Track referrals and enquiries alongside observations. The SEO value guide helps connect visibility work to business outcomes.
Avoid common audit mistakes
Do not score every mention equally. Do not merge ChatGPT, Perplexity, Google and Bing results into one unexplained ranking. Do not estimate market share from a few manual prompts. Do not copy a rival’s claims without verifying them. An honest audit should show the sample and its limits.
Compare the same decision across platforms
An audit should not assume that all AI products show sources in the same way. Keep separate worksheets for ChatGPT, Perplexity, Google AI Overviews or AI Mode, and Bing’s supported AI experiences. Record what each platform actually displays. A page cited in Bing may never appear in your ChatGPT sample, and that difference is worth noting without turning it into a universal ranking.
Use a consistent question set within each platform. For example, a Hong Kong agency might test how to choose a provider, what a service includes and what affects its price. A clinic might test appointment preparation and differences between treatment options, with professional review of any medical interpretations. Avoid questions designed only to force the brand name into the answer. They will overstate real discovery.
Where an answer gives no citation, record that fact. Where it cites a directory, inspect the directory entry and the underlying service page. Where it cites a competitor’s article, examine whether the article answers a distinct question or provides evidence your site lacks. The audit is a diagnostic exercise, not a competition for screenshots.
Keep the sample honest
A sample of twenty questions does not represent every prompt a person might ask. Results can vary with location, language, session context and platform updates. State those limits in any client report. If you alter the wording in the next audit, mark the question as new instead of claiming a direct trend.
Record the answer in context rather than only whether a brand appeared. A brand might be mentioned as an example of a category, as a provider to consider or as a warning. These meanings differ. Also distinguish a citation to a company page from a citation to an independent article that merely names the company.
For a useful business decision, connect the audit to site data. If a competitor is repeatedly cited for a topic that brings qualified enquiries, improving your own answer has a clear rationale. If a topic attracts attention but no relevant customers, its priority may be lower. Ask the sales team whether prospects bring up the issue. The goal is to direct editorial effort where it can help a real buyer.
A good audit ends with a short, owned action list. It does not end with a promise to “dominate AI” based on a small sample.
Use findings to improve the right asset
Suppose competitors appear for a pricing question. If your pricing page already exists but hides all useful information, improve the explanation of scope and fee factors. A new blog post may only compete with the service page. If no page explains a distinct comparison customers repeatedly ask for, a new article may be justified.
Record the intended action before writing: improve an existing URL, create a new page, correct a public listing or leave the subject alone. Then assign a way to judge progress. This prevents a competitor audit from becoming a list of screenshots that creates content without a clear reader need.
The audit can also reveal that a competitor is present for a question outside its actual service. Do not assume that appearance is worth copying. A business should prioritise queries where it can deliver the promised service and where a reader has a sensible next step. Keep this commercial filter visible when selecting future content.
Where possible, ask a colleague to review the action list without seeing the brand scores. If the recommended work still helps customers, the audit has probably identified a genuine opportunity.
A clear owner and follow up date are essential, or the same unresolved finding will reappear in the next report.
FAQs About AI Competitor Audits
How many questions should an audit include?
Use enough to cover the main customer decisions, but keep the set small enough to repeat. A carefully chosen set is more useful than hundreds of generic prompts.
Is a competitor citation evidence that its SEO is better?
Not by itself. A citation shows what appeared in that answer at that time. Review the source page and broader business results before drawing a conclusion.
Should we copy cited competitors’ article structures?
No. Learn what questions they answer well, then provide a better answer grounded in your own expertise and evidence.
How often should the audit be repeated?
A monthly or quarterly schedule is usually easier to interpret than constant checking. Record major content and platform changes between audits.
