What is a good AI visibility score?

What does the whole distribution look like?

Bunched low, with a thin tail at the top.

Half of the 300 businesses we scanned landed between 20 and 39. The lowest score was 4 and the highest 90. Only 50 of the 300 reached 60, and only 12 reached 80.

What does each band mean, and why are ours 70 and 35?

70 and above we call Visible, 35 to 69 Partly visible, below that Weak or Not found. Those are our thresholds, not an industry standard.

They are set where they are for a reason you can check against the same dataset:

• 70 is the 90th percentile of the Australian businesses we scanned. Visible does not mean adequate. It means the top tenth of a sample that was already drawn from businesses ranking well in classic search.
• 35 sits just under the median of 36. Below it you are in the bottom half of a sample of businesses that are, by construction, doing better than average online.
• Below 35 the band depends on recognition. A business the engines know but never volunteer scores as Weak. One they cannot place at all scores as Not found. That distinction matters, because the two problems have different fixes.

Nobody has standardised any of this. Another vendor's 70 is not our 70, because it is a different formula, on different questions, on a different day. A score is only comparable against scores from the same tool. Ours is published in full on The Receipts Engine: three quarters of the overall score is what the engines said, one quarter is site readiness, and within each engine 70% is a straight count of the questions you were named in, 20% is recognition and 10% is sentiment.

How does my industry change the answer?

A lot. A 45 is mid-table for an e-commerce brand and close to top-quartile for a tradie.

So which number should I compare myself against?

Your own industry's median first, then its 75th percentile as the target.

The national mean of 38.3 is a headline, not a benchmark. An accountant on 38 is above their industry's median of 29 and roughly at its 75th percentile. An e-commerce brand on 38 is well below its industry's median of 55.5. Same score, opposite verdicts.

One more comparison worth having in mind: the 36 recognisable national anchor brands in the sample averaged 52.5 out of 100 against 36.4 for the other 264 businesses. A famous name is worth roughly 16 points on this scale, and it still guarantees nothing. Three household names in our sample, 1300SMILES Dentists, Australian Skin Clinics and Turner Freeman Lawyers, were not named by Gemini in any of our test questions.

Source: Answerable, The State of AI Visibility in Australia 2026. 300 Australian businesses, six industries, scanned 30 July 2026. Raw data published as CSV and JSON under CC BY 4.0.

Why is a score a measurement and not a forecast?

Because it records what two engines said on one day, to one fixed set of questions.

• Each question was asked once. There were no repetition runs, so a business one answer short of a mention on the day stays unmentioned in the data.
• The engines disagree with each other constantly. Across the 137 businesses with complete records from both, Gemini scored the same business higher than ChatGPT for 105 of them, ChatGPT was higher for 7, and 25 were tied. Across all 137 the two engines' scores differed by an average of 25.3 points out of 100. Same business, same day, a quarter of the scale apart.
• The models change. This is a snapshot from one scan week in July 2026, and we publish it as one rather than as a trend we do not yet have.

The practical consequence: read a single score as a starting position, and read the direction over several months as the actual signal. A one-point move between two scans is noise.

What can a score not tell you?

Whether anyone asked, whether being named made a sale, and what it will be next month.

• Demand. We have no query volume data and no access to anyone's real ChatGPT or Gemini conversations. We ask a fixed set of questions ourselves through public APIs and count what comes back.
• Attribution. Being named in an answer does not produce a referral you can find in analytics, which is its own problem: see how to track ChatGPT traffic in Google Analytics.
• Causation. The score will not tell you which of your gaps caused which absence. That takes reading the answers themselves.
• Comparability. Not against another tool's score, and not against a score from a different question set.

Get your own number, free.

A benchmark is only useful next to your own score. The free scan gives you one, plus the questions we asked and who was named instead of you.