Most sites blocking AI training crawlers are blocking the search ones too
56.2%
Of the sites in our corpus that block GPTBot, 56.2% also block OAI-SearchBot, which removes them from ChatGPT's search answers.
Training vs search crawlers · August 2026
AI visibility statistics · primary data · free to cite
Of more than a million websites, 216 are ready for AI search.
The headline statistics from SearchScore's audit data, drawn from a live corpus of over 1,000,000 website audits and including the published State of AI Visibility Index (SAVI), in one citable page. Every number below is a self-contained, sourced statistic you can quote, with a stable anchor link and its sample size and date stated. Updated August 2026 with the Q3 2026 SAVI. Next update: with the Q4 2026 edition.
1M+
websites analysed
216
AI-Ready (GEO 80+), about 1 in 4,600
0
Verified (GEO 90+)
86.6
highest score, Q3 2026 census
Source: Q3 2026 SAVI Report, published index values.
✓ 1,000,000+ sites audited ✓ 300+ signals per site ✓ Free to cite with attribution
The whole picture, in one dot
About one website in 4,600 is AI-Ready. #
Every dot is a website. The lit one is ready to be recommended.
4,600 dots, 1 lit. The published AI-Ready rate: 216 of 1,000,000+, 0.022%, about 1 in 4,600.Source: Q3 2026 SAVI
The ceiling
At the Q3 2026 census, nobody had cleared the bar. #
In the Q3 2026 census the highest score was 86.6, 3.4 points short of Verified (90+), which no site had reached.
"The top of the market is unoccupied. An average tells you where the middle of the web sits. A ceiling tells you what is achievable with the way sites are built today, and this one falls short of the bar."Q3 2026 SAVI Report, Nobody has cleared the bar
Source: Q3 2026 SAVI Report (census figures; 3.4 is 90 minus 86.6). Tier bands per the methodology: AI-Ready 80 to 100, Strong 70 to 79, Emerging 40 to 69, Low Visibility 20 to 39, Invisible 0 to 19; Verified is a gold super-tier at 90 and above.
The statistics
Ten numbers, each with its sample and date.
63.8% of scored audits fall below 50 for AI visibility #
63.8%
63.8% of the scored audits in SearchScore's live corpus (July 2026) fall below 50 out of 100 for AI search visibility (GEO score).
Source: the SearchScore live audit corpus, July 2026. Context: What is GEO? · How it is measured: Methodology
Only 216 of over a million analysed sites are AI-Ready (GEO 80+) #
0.022%
Only 216 of the 1,000,000+ sites we've analysed are AI-Ready, a GEO score of 80 or above: about 1 in 4,600 (July 2026).
Source: the SearchScore live audit corpus, July 2026. Context: What is GEO? · How it is measured: Methodology
72.4% of websites have no llms.txt file #
72.4%
In the Q3 2026 SAVI signal sample, a 20,000-site sample of the most recent audits scored on the current engine, 72.4% of websites have no llms.txt file, the proposed standard for describing a site's key content directly to AI crawlers.
Source: the AI-Ready ceiling study, August 2026, from the Q3 2026 SAVI signal sample. Context: llms.txt examples · How it is measured: Methodology
13.5% of websites block GPTBot #
13.5%
Across 347,988 websites audited by SearchScore (August 2026), GPTBot was blocked on 13.5% while OAI-SearchBot was blocked on 7.83%. Of GPTBot blockers, 42.6% allow OpenAI search.
Source: AI crawler policy 2026, N = 347,988, August 2026. Context: What is GEO? · How it is measured: Methodology
68.2% of websites have no Organisation schema #
68.2%
Across the Q3 2026 SAVI index panel of 102,873 websites, weighted, 68.2% of websites publish no Organisation schema, the structured data AI engines use to identify who is behind a website when deciding what to cite. 55.7% publish no JSON-LD structured data at all. This is the same panel and weighting as the 45.24 index mean.
Source: Q3 2026 SAVI index panel, 102,873 websites, weighted, Q3 2026 SAVI Report. Context: Structured data for AI engines · How it is measured: Methodology
216 of 1,000,000+ analysed sites are AI-Ready, and none were Verified at Q3 2026 #
216/1,000,000+
In the Q3 2026 State of AI Visibility Index, 216 of more than 1,000,000 analysed websites score AI-Ready (GEO 80 or above). That is 0.022%, about 1 in 4,600. In the Q3 2026 census no site had reached the Verified tier (GEO 90 or above); the highest score was 86.6.
Source: published index values, Q3 2026 SAVI Report · Index home: the SAVI Index
The fixed index panel reads a weighted mean GEO of 45.24 at wave 1 #
45.24/100
From Q3 2026 the index is built on a fixed panel of 102,873 domains, re-audited each wave on the same method and scorer version, so that movement between editions reflects the web rather than changes in the sample. Wave 1 is dated 1 August 2026. Earlier editions were scored on a different engine and on a corpus that more than doubled between them, so Q3 publishes no like-for-like average comparison with Q1 or Q2.
Source: SAVI index panel, wave 1, Q3 2026 SAVI Report
Just over half of websites land in the Q3 Emerging band (40 to 59) #
52.67%
The Q3 2026 SAVI tier distribution, across a population-weighted sample of the 102,873-site index panel. These are the band definitions the Q3 edition used. The current tier scale sets Strong at 70 to 79 and Emerging at 40 to 69, so these shares describe the Q3 bands only.
- Q3 2026 bands:
- AI-Ready (80 to 100) 0.03%
- Strong (60 to 79) 13.14%
- Emerging (40 to 59) 52.67%
- Low Visibility (20 to 39) 32.28%
- Invisible (0 to 19) 1.89%
Source: Q3 2026 SAVI Report · Current tier scale: Methodology
SearchScore has audited more than 1,000,000 websites #
1M+
SearchScore's live audit corpus covers more than 1,000,000 website audits for AI search visibility, each scored against 150+ AI visibility signals weighted across 8 GEO categories, and grows daily. The published Q2 2026 SAVI Report was built on the 850,000-audit panel it reached at that time.
Source: live-corpus count from the SearchScore audit database, July 2026 (1,000,000+ distinct domains scored in the trailing window); the 850,000 figure is the published corpus of the Q2 2026 SAVI Report. How it is measured: Methodology
Greater London accountancy firms average 52.8/100 for AI visibility #
52.8/100
SearchScore's June 2026 benchmark of more than 150 Greater London accountancy firms found an average AI visibility (GEO) score of 52.8 out of 100, with only four firms scoring above 70.
Sector by sector
Ten sector and city benchmarks from the SAVI editions. #
Each SAVI sector edition audited a defined group of businesses and published its average AI visibility (GEO) score and how many reached AI-Ready (80+). These are the six sector editions' own headlines.
26,204
UK businesses audited across ten editions. Thirteen are ready to be recommended.
Sum of the ten editions' audited counts (1,567 + 1,038 + 3,399 + 151 + 1,977 + 9,487 + 2,307 + 100 + 100 + 6,078). AI-Ready counts from the seven editions that state one (1 + 1 + 2 + 0 + 3 + 0 + 6); the three accountancy city editions state none, so thirteen is a floor. Each edition was scored on the engine of its time.
3AI-Ready of 9,487
Britain has just three law firms AI is ready to recommend.
UK Law Firms 2026 · published headline
2AI-Ready of 3,399
Britain has just two dental practices AI is ready to recommend.
UK Dentists 2026 · published headline
1AI-Ready of 1,038
Britain has one accountancy firm that AI is ready to recommend.
UK Accountants 2026 · published headline
1AI-Ready of 1,567
Britain has just one aesthetic clinic AI is ready to recommend.
UK Aesthetic Clinics 2026 · published headline
0AI-Ready of 1,977
Britain has not one vet AI is ready to recommend.
UK Veterinary Practices 2026 · published headline
0AI-Ready of 2,307
Britain's best-rated care homes are the ones AI cannot find.
UK Care Homes 2026 · published headline
Every sector average on the same scale. Not one reaches Strong.
All nine published averages sit between 46.4 and 56. Strong begins at 70 on the current tier scale.
| Edition | Audited | Average GEO | AI-Ready (80+) | Published |
|---|---|---|---|---|
| UK aesthetic clinics | 1,567 | 56 | 1 | July 2026 |
| Birmingham businesses | 6,078 | 54.9 | 6 | October 2026 |
| UK accountancy firms | 1,038 | 53 | 1 | June 2026 |
| UK dental practices | 3,399 | 53 | 2 | July 2026 |
| Greater London accountants | 151 | 52.8 | Not stated | June 2026 |
| UK veterinary practices | 1,977 | 52 | 0 | July 2026 |
| UK law firms | 9,487 | 51.4 | 3 | July 2026 |
| UK care homes | 2,307 | 48.3 | 0 | July 2026 |
| Birmingham accountants | 100 | 47 | Not stated | May 2026 |
| Manchester accountants | 100 | 46.4 | Not stated | May 2026 |
Source: each linked SAVI edition. Every edition is a dated snapshot scored on the engine of its time, so compare firms within an edition rather than editions with each other. The city editions do not state an AI-Ready count.
Where sites fall short
Reachable, but not quotable. #
The GEO score is built from eight weighted categories. In the Q3 2026 SAVI, sites scored well on the plumbing an AI needs to reach them and poorly on what it needs to trust and cite them. Sites pass the plumbing and fail the trust test: Technical averages 70.8 and Brand Authority 23.7, a 47-point gap.
"Note which category carries the most weight. E-E-A-T Content is 24% of the score, the heaviest of the eight, and it averages 38.5."Q3 2026 SAVI Report
Source: average score per category out of 100, from a 20,000-site sample of the most recent audits scored on the current engine, Q3 2026 SAVI Report. Weights are the GEO category weights per the methodology.
From the research
The findings people argue with. #
Perplexity shows its sources 10x more than any other AI engine
73.8%
Across 10,201 source URLs collected from six AI engines, 73.8% came from Perplexity alone. ChatGPT returned 4.2%, Gemini 3.8%.
AI source disclosure · August 2026
llms.txt compliance: what a 4,000-file crawl found
41.8%
In a 4,000-file crawl, 41.8% of sites that appear to have an llms.txt serve HTML at that path. Of the files that genuinely exist, 79.4% pass.
llms.txt compliance crawl · August 2026
Does llms.txt work? What 102,873 websites show
+5.36
Across the 102,873-site index panel, sites with an llms.txt score 5.36 points higher, and a quarter of that is the scorer counting the file. A pre-registered result.
Does llms.txt work? · August 2026
Britain's best-rated care homes are the ones AI cannot find
0 of 2,307
We checked 2,307 UK care home websites to see how ready they are for AI search. Not one is AI-Ready, and CQC Outstanding homes are no easier to find than the rest.
UK care homes study · July 2026
Organisation schema is a marker, not a lever
176 of 178
176 of the 178 AI-Ready websites in the 102,873-site index panel publish Organisation schema, but the tag itself is worth 1.45 points. A pre-registered result.
Organisation schema · August 2026
Sector deep dives: UK law firms · UK care homes. Cross-edition analysis: AI visibility benchmark 2026 · the AI-Ready ceiling. All studies: Research.
Place your score
Where would your site sit? #
Type a GEO score to see its tier. Your real score takes 60 seconds on the free audit.
Tiers per the methodology: AI-Ready 80 to 100, Strong 70 to 79, Emerging 40 to 69, Low Visibility 20 to 39, Invisible 0 to 19. This names a tier; it is not a percentile.
Get my real scores →Every SAVI edition
The State of AI Visibility Index, edition by edition. #
Each edition is a dated record and keeps the figures it was published with.
Global index
- Vol. 3, Q3 2026: Nobody has cleared the barAugust 2026 · current edition
- Vol. 2, Q2 2026: The AI visibility gap is getting widerApril 2026
- Vol. 1, Q1 2026: The state of AI search visibilityMarch 2026
UK and regional
- UK AI Search Visibility IndexApril 2026
- West Midlands AI Search VisibilityApril 2026
UK sectors
- UK care homesJuly 2026
- UK law firmsJuly 2026
- UK veterinary practicesJuly 2026
- UK aesthetic clinicsJuly 2026
- UK dental practicesJuly 2026
- UK accountancy firmsJune 2026
Cities
- Birmingham businessesOctober 2026
- Greater London accountantsJune 2026
- Manchester accountantsMay 2026
- Birmingham accountantsMay 2026
Data and citation packs
Check the numbers yourself. #
Aggregate data, methodology and citation metadata for the main studies, as JSON. Row-level website and domain lists are not published.
- Q3 2026 SAVIJSON
- AI crawler policy 2026JSON
- AI source disclosure 2026JSON
- UK law firms 2026JSON
- UK care homes 2026JSON
All packs, with licence and citation text: Research data and citation packs.
Methodology note
Where these numbers come from.
The statistics on this page come from two related sources, and each one states which. The published SAVI figures (the 1,000,000+ analysed corpus and the quarterly index values) are report-level numbers from the Q3 2026 SAVI Report. The full scoring methodology and verification rules are published at /savi/methodology/. The per-signal rates are computed directly from SearchScore's live audit database, and each is quoted with its own sample size and date: score-distribution figures cover every scored audit in the live corpus, while signal-level rates come from the published studies each one cites: AI-crawler blocking from a 347,988-site study, llms.txt from the Q3 2026 SAVI signal sample of 20,000 recent audits, and Organisation schema from the Q3 2026 SAVI index panel of 102,873 websites, weighted.
Every audit uses the same automated pipeline that powers the free scan: it fetches the live page, robots.txt, llms.txt and sitemap, checks external authority and review signals, and scores the site across 8 weighted GEO categories. No sites are excluded for being too small or too large, and no curation is applied.
The full measurement approach, including how each signal is collected and verified, is documented on the methodology page. The quarterly index built from this data, with tier distributions and category averages, is the State of AI Visibility Index (SAVI).
Updated August 2026 with the Q3 2026 SAVI. Next update: with the Q4 2026 edition. These statistics are free to cite with attribution and a link. Suggested citation format:
CitationSearchScore (2026). AI Search Statistics. searchscore.io/data/