OpenAI Most Often Named in U.K. Companies’ AI Disclosures, Study Finds

Insider Brief

• OpenAI accounted for 48.7% of 1,377 documented links between U.K. companies and AI providers in GreenPT’s study, followed by Anthropic, open-source models and Google.
• OpenAI and Anthropic together accounted for 72% of the links, while open models appeared across all nine industry groups examined.
• The findings do not establish market share or national adoption rates, and GreenPT estimated that 31% of links involving OpenAI, Anthropic and Google lacked sufficient evidence of use.

OpenAI was the most frequently named AI model provider in public documents linking U.K. companies to the technology, according to a GreenPT study.

GreenPT collected 1,377 documented links between British companies and AI providers from sources including company statements, supplier case studies and privacy notices. OpenAI accounted for 670 of those links, or 48.7%, about twice Anthropic’s total.

Anthropic followed with 322, or 23.4%. A combined category of open-source models ranked third with 235, or 17.1%, ahead of Google with 150, or 10.9%, according to the November report.

Together, OpenAI and Anthropic accounted for 72% of the company-provider links in the main ranking. The findings show how prominently the two suppliers appear in the public evidence GreenPT collected, though they do not establish their shares of customers, spending or actual AI workloads.

GreenPT, a European AI platform that markets services focused on privacy and sustainability, counted each company once for each provider connected to it. A business naming both OpenAI and Anthropic therefore appeared under both. The total of 1,377 does not represent 1,377 separate businesses.

The study also found that a provider’s appearance in a document did not always establish that its technology was being used. GreenPT estimated that about 31% of the links involving OpenAI, Anthropic and Google lacked evidence meeting its standard for corroborated use.

Some involved proposed projects, offered services, outdated statements or conditional language. Others described pilots or plans rather than established use. The distinction limits what readers can infer from the rankings.

What the Documents Show

GreenPT reviewed publicly available material across industries, including disclosures from suppliers and their customers, privacy notices and lists identifying companies that process data for a business.

Each counted entry linked one U.K. company to one AI model provider. These entries, which the report calls references, formed the basis of the comparison.

The research was a review of documents rather than a representative survey of British businesses. Companies that had not publicly discussed their AI suppliers were absent, so the findings cannot establish how widely AI is used across the country.

To conduct the comparison, GreenPT separated its research into two datasets. The main ranking contained the totals for OpenAI, Anthropic, Google and the combined open-source category. A second dataset assessed the evidence behind 734 company-provider links across 15 providers.

In that second dataset, GreenPT classified an entry as supported when it found disclosed use or an explicit identification of an AI supplier, along with a corroborated connection to a U.K. company. A statement from the company or its supplier was sufficient; GreenPT did not independently check whether the technology was running inside the business.

Other entries had weaker or different evidence. Some were conditional, outdated or unclear about the company’s identity. Others concerned offered integrations, pilots, plans, research, historical use or an attribution the researchers rejected.

The proportion meeting GreenPT’s standard was similar for the three major named providers: 69.9% for OpenAI, 68.5% for Anthropic and 65.7% for Google.

GreenPT said those differences were within the statistical uncertainty of the samples. OpenAI’s larger total therefore distinguished it more clearly than the strength of the evidence behind each documented connection.

The researchers then applied those percentages to the larger totals in the main ranking. Combining those estimates with the open-source results, GreenPT estimated that 1,023 of the 1,377 links, or 74.3%, met its evidence standard.

That figure is a projection, as the researchers did not individually classify every entry in the main ranking. The report also acknowledges that extending the smaller dataset’s results to the larger one requires an assumption.

Open Models Appear Across Industries

The open-source category accounted for all 235 of its documented links meeting GreenPT’s standard. That result raised the estimated evidence-backed share of the overall dataset above the roughly 69% recorded for OpenAI, Anthropic and Google combined.

GreenPT’s open-source category covers open-weight models used directly or through services that send requests to different models. Weights are the numerical settings a model learns during training. Making them available allows others to run and adapt the model.

The category combines multiple models, while the other main ranking entries identify individual suppliers. Its third-place position consequently reflects the standing of a group of technologies rather than a single company.

GreenPT also examined differences among industries using a separate set of 969 entries. That analysis combined the 734 entries assessed for evidence with the 235 open-source entries.

Software, IT and AI services accounted for 392 entries, about 40% of that industry dataset, including 278 that met GreenPT’s standard. Financial services followed with 138 entries, including 109 supported by the required evidence. Business and professional services had 95, of which 81 met the standard.

Education had the highest proportion backed by sufficient evidence, at 93.7%, or 74 of 79 entries. Marketing, media and creative businesses had the lowest, at 63.5%, or 47 of 74.

Open models accounted for between 28% and 33% of evidence-backed entries in each of the nine industry groups. OpenAI led the named providers in most groups, but Anthropic led in retail, food and hospitality, with 16 entries meeting the evidence standard compared with OpenAI’s 10.

Anthropic tied OpenAI in marketing, media and creative businesses and in healthcare and life sciences. Google’s strongest industry showing was financial services, where it accounted for about 16% of evidence-backed entries.

GreenPT cautioned that smaller groups, including healthcare with 31 total entries and legal services with 47, were too small to support firm comparisons among providers.

The main ranking also leaves out some suppliers included elsewhere in the research.

Twelve additional providers appeared in the evidence-assessment dataset, including Mistral, Amazon, IBM, xAI, Meta, Alibaba’s Qwen, Cohere, Stability AI, DeepSeek, Midjourney, Stable Diffusion and Voyage AI.

Together, they accounted for 81 entries, of which 60 met GreenPT’s evidence standard. The report said adding those entries to the main ranking would increase the total to 1,458 and reduce OpenAI’s share to 46%, Anthropic’s to 22.1%, open source’s to 16.1% and Google’s to 10.3%.

All 25 Amazon entries met the standard. GreenPT identified Amazon as the only smaller provider whose statistical range clearly exceeded the combined evidence-support rate of OpenAI, Anthropic and Google. Most smaller suppliers had too few entries for reliable comparisons.