Original research
The State of AI Visibility in Australia 2026: who AI recommends, and who it can't see.
We scanned 300 of Australia's most search-visible businesses across six industries on 30 July 2026 and counted who ChatGPT and Gemini actually name when a buyer asks.
Key findings
Eight numbers that describe the state of play
Every figure below comes straight from the dataset published at the bottom of this page. Nothing is modelled, projected or annualised, and how to read each number is spelled out further down.
The average Australian business in our sample scores 38.3 out of 100 for AI visibility. The median is 36.
41% of the 137 businesses with complete records from both engines were never named once by either ChatGPT or Gemini, across every question we asked.
26% of those same 137 are recognised by AI when asked about directly, yet never named when a buyer asks who to use.
76% of the tradie businesses with complete two-engine records were never named by either engine, the worst result of the six industries.
Websites average 74.9 out of 100 for on-site AI readiness, yet readiness barely tracks the per-engine visibility scores: Pearson r = 0.116 (n = 137).
13% block AI crawlers outright in robots.txt, so the engines cannot read their websites at all.
71% have no llms.txt file. We report it as a legibility signal only: Google has said it does not use llms.txt.
Gemini's visibility score for the same business was higher than ChatGPT's 77% of the time across the 137 complete two-engine records, with an average gap of 25.3 points between the engines.
Household names are among the invisible. The 36 recognisable anchor brands in the sample average 52.5 out of 100 against 36.4 for everyone else, yet 1300SMILES Dentists, Australian Skin Clinics and Turner Freeman Lawyers were not named by Gemini in any of our test questions. That is a statement about our questions on one day, not about those businesses. It is also how little a famous name guarantees on its own.
By industry
Six industries, one pattern
Fifty businesses per industry. Mean score is the average AI visibility score out of 100. "Never named" means zero mentions from both engines across every question asked, and it is counted over the businesses in each industry with complete records from both engines (the "Complete records" column), because blank Gemini rows cannot be counted either way. Crawler blocking is counted over all fifty.
| Industry | Businesses | Complete records | Mean score | Never named by either engine | Blocking AI crawlers | Best performer |
|---|---|---|---|---|---|---|
| E-commerce & retail | 50 | 22 | 55 | 9% | 12% | Bondi Sands (89) |
| Dentists | 50 | 29 | 38.5 | 41% | 14% | Pacific Smiles Dental (76) |
| Law firms | 50 | 23 | 37.5 | 30% | 12% | Slater and Gordon Lawyers (85) |
| Cosmetic & skin clinics | 50 | 23 | 35.2 | 43% | 16% | Laser Skin & Vein Clinic (78) |
| Accountants & advisers | 50 | 19 | 34 | 47% | 12% | Etax Accountants (90) |
| Tradies | 50 | 21 | 29.8 | 76% | 12% | Fallon Solutions (71) |
| All industries | 300 | 137 | 38.3 | 41% | 13% | Etax Accountants (90) |
Methodology
How to read these numbers
This is a snapshot from a single scan week, with every row dated 30 July 2026. Here is exactly what we did, and exactly where the numbers stop being able to tell you things.
The sample
300 businesses, 50 per industry, across accountants, cosmetic and skin clinics, dentists, e-commerce, law firms and tradies. For the local-service niches, businesses were drawn from five mainland capital cities. The sample frame was the top organic search results for buyer-style queries, plus a set of recognisable national anchor brands in each industry.
That frame matters. Search-based sampling skews toward businesses that already rank well in classic search, which strengthens the finding rather than weakening it: the businesses that AI never names are drawn from the visible end of the Australian market, not the invisible end.
The questions
The ChatGPT signal comes from five shared buyer questions per industry-city cohort, the kind a real customer would ask, such as who to use for a service in a given city. The Gemini signal comes from the same sweep plus one grounded direct question per business. 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 this data.
Named versus recognised
A business is named when an engine volunteers it in answer to a buyer question. It is recognised when the engine shows it knows the business exists once asked about it directly, but never volunteers it. The 26% figure, counted over the 137 businesses with complete records from both engines, is that second group: real businesses the models know about and do not recommend.
The scores
Each business carries several scores in the dataset. The AI visibility score (out of 100) summarises its whole scan record across both engines, site readiness included. The per-engine visibility scores (chatgpt_visibility_score and gemini_visibility_score, out of 100) record how visible the business was to each engine separately, and they are the fields behind the correlation and engine-comparison figures on this page. The site readiness score (out of 100) comes from automated on-site checks: whether AI crawlers are allowed in robots.txt, whether structured data (JSON-LD) is present, whether meta basics pass, and whether an llms.txt file exists. On llms.txt, be careful with the hype: we measured adoption because it says something about how legible a site is trying to be to machines, but Google has said it does not use the file, and nobody should sell it as a ranking lever.
Across the 137 businesses with complete records from both engines, the correlation between site readiness and the mean of the two per-engine visibility scores, both published in the dataset, is Pearson r = 0.116. Readiness explains almost nothing about who gets recommended. Correlate site readiness against the overall AI visibility score instead and you will get a much larger number, because that score folds site readiness into itself; the per-engine columns are the fields this figure is computed on.
The caveats, plainly
- One scan week. These models change their answers, so this is a snapshot, not a trend.
- Five buyer questions per cohort on ChatGPT, plus one grounded direct question per business on Gemini, each asked once. A different phrasing on a different day would produce different individual rows, which is why we publish every row.
- Businesses are named factually. "Was not named in our test questions" describes our questions on one day, not the quality of anyone's work.
- Complete two-engine records exist for 137 of the 300 businesses. The headline never-named and recognised-but-never-named percentages, the industry table's never-named column, the engine comparison and the readiness correlation are all computed on those 137. Blank Gemini fields in the raw data mark the rows without a complete Gemini record.
If the vocabulary here is new, our plain-English guides to AI visibility and AEO, SEO and GEO cover it without the jargon.
What it means
What actually moves the needle
The most useful thing in this dataset is a gap: websites average 74.9 out of 100 for AI readiness while the businesses behind them average 38.3 for AI visibility, and readiness barely correlates with the per-engine visibility scores (r = 0.116, n = 137).
Read that gap honestly and the conclusion is not complicated. On-page readiness is necessary, and 13% of businesses fail even that bar by blocking AI crawlers outright. But it is nowhere near sufficient. The engines are not choosing businesses because their websites are tidy. They are choosing businesses the rest of the web already talks about: reviews, directories, press, industry lists, other people's pages. The anchor-brand numbers point the same way, and so does the fact that three household names still went unnamed by Gemini. A footprint helps and nothing guarantees.
So fix the site, then earn the mentions. If you want to know where you stand first, the same scan we ran on these 300 businesses is free.
Raw data
Take the data
All 300 rows, one business per row, every field we recorded. Licensed CC BY 4.0 with attribution to Answerable. Anyone can re-run our numbers or challenge them, and we would genuinely like to hear about it if you get a different answer. If you just want the headline figures with citable anchors, they are collected on our AI visibility statistics page.
Questions
Common questions about this study
How many Australian businesses are invisible to AI?
Does a good website get you recommended by ChatGPT or Gemini?
Which industry is most invisible to AI?
Do ChatGPT and Gemini agree on who to recommend?
Can I reuse this data?
Find out whether AI names you at all.
The same scan we ran on these 300 businesses is free for yours. See whether ChatGPT and Gemini name you, who they name instead, and you keep the report.