How to Read an AI-Generated Report Without Getting Fooled

The Fact, Inference, or Guess Framework

The Three Categories

Every statement AI makes about your business falls into one of three buckets:

1. Fact.

Verifiable. Either true or not. You can check it.

Go look at your homepage. It either says that or it doesn’t. That’s a fact.

No supporting evidence. Presented as analysis.

The AI has no ranking data. It has no idea who’s outranking whom. This is a guess wearing the clothes of a finding.

Here’s an exercise. Below is a sample paragraph from an AI-generated website audit. Read it and label each claim:

Let’s break it down:

• "Clean, professional design." Inference. Subjective assessment based on what the AI sees. Reasonable, but not measurable.

• "Appeals to families in the suburban Denver market." Guess. The AI has no customer data. It’s associating design cues with a demographic it’s assuming.

• "Brake service page lacks depth compared to competitors." Could be a fact or a guess, depending on whether the AI actually read your page and your competitors’ pages. Ask it: did you access these URLs?

• "Google ranking has likely declined." Guess. The AI has no ranking data. The word "likely" is your signal.

• "Adding more content targeting long-tail keywords." Generic inference. Not wrong as general advice, but it’s not based on your actual search performance or keyword gaps.

One paragraph. Five claims. One possible fact, one inference, three guesses. And the whole thing reads like a professional audit.

Certain words are reliable indicators that the AI is guessing or inferring rather than reporting:

• "Likely"
• "Typically"
• "Based on the area"
• "Potentially"
• "It’s common for"
• "Shops like yours usually"
• "In most cases"

When you see these words, you’re reading an assumption. The assumption might be reasonable. But it’s important to know you’re looking at an estimate, not a verified finding.

Also watch for confident-sounding statements with no source. “Your competitors are investing heavily in paid search” sounds specific. But the AI has no data on your competitors’ ad spend. It’s filling a gap with a plausible-sounding claim.

This is the one that gets almost everyone.

AI output that’s formatted with bullet points, numbered lists, headers, bold text, and structured sections feels more credible than the same information in a plain paragraph. That’s a design feature of how these tools present information. It is not an indicator of accuracy.

A perfectly formatted list of recommendations can be entirely wrong.

Formatting is not evidence.

Four steps you can use every time you get AI output:

1. Ask for sources. “For each factual claim you made, can you provide a specific source or URL?” The AI will either give you real links, give you links that don’t exist (hallucinated URLs), or admit it doesn’t have a source. All three answers are useful.
2. Click every link. If the AI provides sources, open them. Verify the page exists, that it says what the AI claims, and that it’s credible. AI tools are known to generate plausible-looking URLs that lead nowhere.
3. Test one verifiable claim. Pick one thing from the output that you can independently check. A ranking claim, a demographic number, a technical assessment. Check it with a real tool or real data. If that one claim is wrong, treat everything else in the output with serious skepticism.
4. Ask follow-up questions. “How do you know that?” and “Is that based on data you have access to, or are you estimating?” are simple prompts that force the AI to show its work. It will often acknowledge its limitations when asked directly.

Shop owners are getting AI-generated audits from coaches, from peer groups, from competitive shops, and from their own curiosity. These audits look professional. They read like expert analysis. And some of the recommendations are solid.

But some aren’t. And you can’t tell which is which without a method for evaluating what you’re reading.

This is this method. Fact, inference, or guess. Five minutes. Every time.

For the complete framework, including sample prompts with built-in verification constraints, download The Shop Owner’s Guide to AI in Marketing. Chapter 6 goes deeper into everything in this post.

Next in the series: a topic we’re hearing about constantly. When a someone (a vendor, a colleague, a coach) says “we use AI,” what does that actually mean? There’s a big difference between ChatGPT (the tool you use on your phone) and the AI that powers features inside platforms like your CRM or marketing tools. We’ll explain what that difference is and why it matters for your business.

If you’re new to the series:

Post 1: AI Is a Flashlight, Not a Map

Post 2: What AI Gets Wrong About Your Shop

Post 3: 5 AI Prompts That Actually Help Your Shop

Post 4: Does Google Penalize AI Content?

For the full framework: The Shop Owner’s Guide to AI in Marketing

Heather Myers is the Chief Technology Officer at KUKUI, where she builds marketing and customer engagement technology for independent auto repair shops. Before joining the automotive technology space, she built information systems for public and academic libraries.

This is the fifth post in our ongoing series, AI Is a Flashlight, Not a Map. New posts publish every two weeks.