The Brand Prompt Audit: Scoring What ChatGPT, Gemini and Copilot Say About You

WhatsApp Channel Join Now

Somewhere this week, a potential customer asked an AI assistant about your business. Maybe they asked for a recommendation, maybe they asked if you’re any good, or maybe they asked what you charge. You weren’t in the room, and unless you’ve checked, you don’t know what it told them.

Many businesses haven’t looked. They track their Google rankings and read their reviews, but the answers AI tools give about them go unchecked, even though those answers can be out of date, incomplete or simply wrong.

Running the same prompts through several AI assistants shows how differently each one describes a business.

Alt text: A laptop screen showing three AI chat assistant windows side by side

A brand prompt audit fixes that. It’s a set of questions you put to the main AI tools, a simple way to score the answers, and a short list of fixes for the gaps you find. It takes an afternoon the first time and about an hour each quarter after that.

Which AI Tools to Test

You can’t test them all, so start with the ones your customers use. Roy Morgan’s March 2026 survey of AI tool usage found that 13.6 million Australians aged 14 and over, or 58% of that group, used an AI tool in an average four weeks.

AI toolAustralian users (March quarter 2026)Share of Australians 14+
ChatGPT10.5 million45%
Google Gemini5 million21%
Microsoft Copilot4 million17%
Claude777,0003%

Source: Roy Morgan, AI tools usage, March quarter 2026.

ChatGPT, Gemini and Copilot cover the large majority of users, so they’re the three to audit. If you sell to a technical audience, add Perplexity and Claude.

Use a clean session

AI tools personalise answers based on your history and location. Log out, or use a private browser window, so you see something closer to what a stranger would see. If your customers are local, include your suburb in the prompt rather than relying on the tool to guess.

The Ten Prompts to Run

Write the prompts the way a customer would, not the way a marketer would. These ten cover the main stages of a buying decision:

  1. “Who are the top-rated [your service] providers in [your suburb or city]?”
  2. “Is [your business name] any good?”
  3. “What does [your business name] do?”
  4. “How much does [your business name] charge?”
  5. “What do customers say about [your business name]?”
  6. “[Your business name] vs [main competitor]: which is better?”
  7. “Where is [your business name] located and what are its opening hours?”
  8. “Who owns [your business name]?”
  9. “I need [a specific job you do]. Who should I call in [area]?”
  10. “What are the alternatives to [your business name]?”

Copy each answer into a spreadsheet along with the date, the tool and the prompt number. A shared spreadsheet also means someone else in the business can run next quarter’s audit without starting from scratch. You’ll want the exact wording later, both to track change and to find where a wrong fact came from.

How to Score the Answers

A score turns a pile of screenshots into something you can compare next quarter. Rate each answer on four points, giving one mark for each.

CheckQuestion to ask of each answerScore
PresenceIs your business named at all?0 or 1
AccuracyAre the facts (services, location, hours, prices) correct?0 or 1
SentimentIs the description neutral or positive?0 or 1
SourceDoes it cite or link to your website or a profile you control?0 or 1

Four marks per answer, so ten prompts across three tools gives a maximum of 120.

A simple scoring grid makes it easy to compare results from one quarter to the next.

Alt text: A clipboard with a printed scoring grid marked with ticks and crosses

Your first score matters less than the pattern. A business that scores well on presence but poorly on accuracy has a different problem from one that isn’t mentioned at all.

What Your Results Usually Mean

You aren’t mentioned for category prompts

If prompts 1 and 9 name competitors and not you, the AI tools don’t have enough independent evidence connecting your business to that service and area. Look at which sources they cite for the competitors. Directories, local news and industry associations often show up, and those are the places to get listed.

The facts are wrong

Wrong hours, an old address or a service you stopped offering usually trace back to an outdated listing somewhere. Start with your Google Business Profile, where Google’s guide to editing your Business Profile walks through updating each field, then check Bing Places, Apple Business Connect and the major directories in your industry.

The tone is negative

If an AI tool describes you unfavourably, it’s usually summarising reviews or forum posts. You can’t delete fair criticism, but you can respond to it publicly, fix the underlying issue, and encourage satisfied customers to leave detailed reviews.

The answer cites someone else’s page about you

Aggregator sites and old news stories sometimes become the main source an AI tool uses for your business. A clear, current “About” page and a service page for each core offer give it a better source to draw from.

Fixing the Gaps

Most fixes fall into three groups, and it helps to tackle them in this order.

Correct the facts first

Inaccurate details can cost you a sale outright, so they come first. Update each profile you control and make sure your name, address, phone number and hours match exactly across them.

Strengthen the signals local search already uses

Google’s own tips to improve your local ranking point to relevance, distance and prominence, and prominence includes reviews and how widely a business is mentioned around the web. The same signals feed many AI answers, so work here pays off twice.

Build the pages the AI tools are missing

If an assistant can’t say what you charge or who you serve, it’s often because your website doesn’t say it clearly either. A plain pricing guide or a “who we work with” page can change the answer within weeks of being indexed.

Tracing a wrong answer back to its source is usually the quickest way to fix it.

Alt text: A brass magnifying glass resting on a printed web page layout

Mistakes That Skew the Results

A few habits make an audit look better or worse than reality. They’re easy to avoid once you know about them.

Asking leading questions

“Why is [your business] the top choice in Newtown?” invites the tool to agree with you. Customers don’t phrase questions that way, so the answer tells you very little. Keep prompts neutral and close to how people actually talk.

Testing once and drawing conclusions

AI tools can give different answers to the same prompt minutes apart. Run each prompt twice and record both answers. If the two disagree, treat the result as uncertain rather than picking the one you prefer.

Ignoring the sources

The cited links are often more useful than the answer itself, because they show where the tool is getting its information. A directory listing you forgot about, or a news story from five years ago, can shape how you’re described today.

Only checking your own name

Prompts that name your business test what the tools know about you. Category prompts, such as “who does commercial electrical work in Alexandria?”, test if you’re considered at all. You need both kinds to see the full picture.

Automating the Check

Running ten prompts across three tools by hand is manageable once a quarter. If you’d like a benchmark before you start, a free AI search score runs a similar scan of how visible a business is across AI assistants and highlights where the gaps are, which gives you a baseline to compare your own audit against.

Make It a Quarterly Habit

AI assistants change the way they answer as their models and search sources update, so a single audit goes stale quickly. Put a recurring reminder in your calendar, keep the same ten prompts, and record the score each time.

Start this week with just the first three prompts in ChatGPT. If the answers surprise you, run the full audit, and you’ll know exactly which of your listings, reviews or pages to fix first

Similar Posts