There are two kinds of business owners right now.
The first type checks their Google Maps ranking and TripAdvisor score and feels reasonably comfortable. The second type has started asking ChatGPT and Gemini "best [their category] in [their city]" — and noticed their competitor's name coming up, not theirs.
If you're in the second group, you've already identified the problem. Now you need a way to measure it precisely, track improvement, and prove that the work is paying off.
That's what the AI Recommendability Score is for.
What is the AI Recommendability Score?
It's a 0–100 number that answers a single question: how likely is AI to recommend your business when a potential customer asks?
It combines six weighted signals that collectively determine AI recommendation likelihood. The score is designed to be stable (repeatable, not subject to daily swings) and actionable (each component points to specific fixes).
Bands:
- 0–40: Invisible — AI rarely or never mentions you
- 41–60: At Risk — you appear occasionally but inconsistently
- 61–80: Visible — AI recommends you in a meaningful share of relevant queries
- 81–100: Leading — you're a top AI recommendation in your category and city
The six components
1. AI Recommendation (20pts)
This is the only component that directly samples AI behaviour. We run your business through 6+ high-intent prompts on ChatGPT (with real web search, the same interface customers use), 3 times each, and measure:
- Mention rate: what percentage of runs did AI mention you?
- Average rank when mentioned: were you #1 or #4 in the response?
This component carries only 20% of the total score weight because AI responses have inherent variability. The other 80% of your score comes from stable, reproducible signals that cause AI to recommend you — so you can improve them without re-running AI queries every day.
2. Review Signal (25pts — the highest weight)
AI models have learned that star ratings are a proxy for quality. Analysis of ChatGPT and Gemini responses consistently shows:
- Businesses below ~4.3★ are systematically excluded from recommendations
- Review volume matters (200 reviews signals more authority than 20)
- Review recency matters (a business with its last review 3 months ago looks inactive)
- Review response rate matters (responding to reviews signals active management)
The Review Signal component scores all four sub-signals and weights them. A business at 4.5★ with 300 reviews and a 90% response rate scores near 100/100 on this component. A business at 4.1★ with 50 reviews and no responses might score 20/100.
The key threshold: 4.3★ is approximately where AI transitions from excluding to including a business in recommendations. Moving from 4.2 to 4.4 can be the single biggest score-mover.
3. Sentiment (15pts)
Star ratings are noisy. A 4.2-star business might have had a difficult quarter and recovered, or might have steady underlying complaints masked by high-scoring ratings from different customer segments.
AI models read the content of reviews, not just the aggregate rating. Recurring complaints about specific issues — wait times, billing errors, staff demeanour — leave a negative sentiment fingerprint that AI picks up on.
The Sentiment component analyses the themes in your recent reviews and scores the positivity/negativity balance. A business with strong stars but recurring complaints about a core issue (e.g. "always slow service") will score lower on Sentiment than its star average suggests.
What to fix: the Sentiment component points directly at operational issues. If sentiment is dragging your score, the fix is operational, not marketing.
4. Listing Consistency (15pts)
AI retrieves information about your business from multiple sources simultaneously — Google Business Profile, TripAdvisor, your website, local directories. If these sources contradict each other (different opening hours, different phone numbers, different cuisine tags), AI reduces its confidence in your listing.
Low confidence = fewer mentions.
Common consistency issues:
- Opening hours that differ between Google and website (especially seasonal or post-COVID updates)
- Address formatting differences (Street vs St, Floor 2 vs 2nd Floor)
- Cuisine or category tags missing on GBP
- Phone numbers with/without country codes differing across platforms
These are easy to fix — and each fix is permanent.
5. Citations (15pts)
AI responses often include source attribution: "according to Time Out," "featured in Dubai's Best Eats 2026," "listed on OpenTable." These citations are drawn from directories, publications and guides that AI models have indexed.
If you're not listed on the directories AI cites for your category and city, you won't get cited — and uncited businesses are less likely to appear in recommendations at all.
This component scores your presence on the key citation sources for your category. The fix is straightforward: identify which directories AI cites and get listed and verified on them.
6. Competitor Gap (10pts)
Your score doesn't exist in isolation. AI compares businesses when generating recommendations. If your top competitor scores significantly higher on review signal, sentiment and consistency, AI will almost always recommend them over you.
The Competitor Gap component benchmarks your score against your top local rivals. A small gap (you're 5 points behind) scores near 100/100 — competitive position. A large gap (you're 30 points behind) scores lower and signals that competitive action is needed.
The value of this component: it turns AI visibility from an absolute measure into a relative one. You might have a solid score of 65/100, but if your competitor is at 82/100, AI will recommend them consistently. The Competitor Gap component makes this visible.
How the score is calculated
The six components are scored 0–100 each, then weighted and combined:
| Component | Weight |
|---|---|
| Review Signal | 25% |
| AI Recommendation | 20% |
| Sentiment | 15% |
| Listing Consistency | 15% |
| Citations | 15% |
| Competitor Gap | 10% |
The resulting score is a single number from 0 to 100. We round to whole numbers — never show false precision.
Stability by design: because 80% of the score comes from stable signals (not live AI sampling), re-running the same audit should produce a score within ±5 points. This means you can trust the score to track real change, not just noise.
Using the score to drive action
The most valuable property of the score is that it points directly at fixes.
A score of 54/100 broken down as:
- AI Recommendation: 35
- Review Signal: 40
- Sentiment: 60
- Listing Consistency: 55
- Citations: 70
- Competitor Gap: 50
...tells you exactly where to focus: Review Signal and AI Recommendation are the lowest. Review Signal has the highest weight. Fix it first. And the specific sub-scores within Review Signal (rating, response rate, recency) tell you exactly how.
This is why the score is more useful than a simple "AI mentioned you 3/10 times" metric. That metric tells you what, but not why or what to do about it.
Tracking score over time
For Fix & Track retainer customers, we re-score weekly and send a Monday morning email with the new score, delta vs last week, and the top fix for this week. Over 8–12 weeks, you build a trend that shows:
- Which components responded to which fixes
- How fast improvement compounds
- Whether you're closing the competitor gap
- Proof of progress for your team, board or franchisees
Get your AI Recommendability Score for free at brandcompanion.ai/ai-visibility. Results in about 1 minute — no login required.