The AI Company Communications Study 2026 benchmarks how 15 AI companies communicate: which usage and revenue numbers they publish, how often, and who speaks for them. Five of the 15 (OpenAI, Anthropic, Google DeepMind, Microsoft AI and Perplexity) disclose concrete active-user or revenue numbers on a regular cadence. The other 10 publish figures only around funding rounds, in aggregates that hide product-level scale, or not at all.
By the Everything-PR Research Team. Data collected February to April 2026. Last reviewed October 9, 2026, with an update section below.
What are the key numbers in the study?
The study's headline numbers are the figures each company disclosed in public between February and April 2026. Each figure carries the date and source type shown in the table.
| Figure | What it measures |
|---|
| 900M | OpenAI ChatGPT weekly active users, February 2026, up from 400M a year earlier |
| $30B | Anthropic annualized revenue, April 2026 |
| 750M | Google Gemini monthly active users in 2026, per Google |
| 1B+ | Meta AI monthly active users in 2026, per Meta |
| 17.8% | Grok share of the US chatbot market, January 2026, third place |
| $200M | Perplexity annualized recurring revenue, September 2025 |
| 8 | Fortune 10 companies that are Claude customers, per Anthropic, February 2026 |
| 5 of 15 | AI companies in the study that disclose concrete usage numbers on a regular cadence |
What were the study's two surprises?
The first surprise is that Anthropic's enterprise revenue passed OpenAI's while its press footprint stayed a fraction of OpenAI's. As of April 2026, Anthropic reported about $30 billion in annualized revenue against about $25 billion for OpenAI. OpenAI's consumer base was about 20 times larger, and its press coverage was several times larger. Communications volume did not track revenue volume.
The second surprise is that the most-covered AI founder carries the lowest documented public favorability. Morning Consult's June 2025 tracking put Elon Musk at 37% favorable and 55% unfavorable. That net-negative sentiment colors how buyers read xAI's communications.
What changed after the study's data was collected?
Anthropic's annualized revenue run rate reached $65 billion at the end of July 2026, up from the $30 billion recorded here for April, according to CNBC's August 17, 2026 report of an investor update. CNBC reported that OpenAI's run rate had reached $40 billion. On October 8, 2026, CNBC reported a figure of about $50 billion for OpenAI, citing an unnamed source.
The $65 billion figure surfaced through an investor update reported by CNBC and Bloomberg, citing sources. It did not come from the regular public cadence this study tracks. Treat every table value below as a snapshot from the collection window, and check the company's latest disclosure before you cite it.
Why does disclosure asymmetry matter to AI buyers?
Disclosure asymmetry matters because buyers compare vendors using numbers that only some vendors publish. OpenAI publishes weekly active users. Google publishes monthly active users. Anthropic publishes enterprise customer counts and revenue but not consumer user counts. xAI publishes little on a regular schedule. Meta reports Meta AI usage together with its wider platforms. Most mid-tier AI companies publish almost nothing about usage.
That gap shapes enterprise buying decisions and press comparisons. For related research on which sources AI engines cite, see the AI Platform Citation Source Index 2026.
How was the study built?
The study draws on public company communications, 2025 and 2026 funding announcements, third-party analytics (Similarweb, Semrush, Apptopia), reporting from Reuters, TechCrunch, The Information, Sacra and SaaStr, and documented enterprise customer disclosures. Every figure is dated to its source. Tier classifications are a qualitative assessment by the Everything-PR Research Team, based on disclosure patterns over the 12 months before publication.
The companies covered are OpenAI, Anthropic, xAI, Google DeepMind and Gemini, Microsoft AI and Copilot, Meta AI, Perplexity, Mistral, Cohere, Character AI, Stability AI, Runway, Midjourney, Inflection AI and Hugging Face. Anthropic and Google DeepMind are both covered, and the figures for each come from public sources, not from the companies.
What three dimensions did the study measure?
| Dimension | Question it answers | Categories used |
|---|
| Implementation messaging | How does the company describe getting its product deployed inside real businesses? | Developer-first (API, documentation, open-model releases), enterprise-first (contracts, cloud distribution, named customer case studies), consumer-first (app store, viral growth, free tier) |
| Usage and adoption disclosure | Which verifiable numbers does the company publish, and how often? | Weekly or monthly active users, seat counts, named customers, revenue, retention, API volume |
| Voice posture | Who speaks in public for the company? | Founder-dominant, dual-voice or corporate-dominant, silent or minimal |
Buyers who want a measurement layer for AI answers can pair this study with the Citation Share Index.
What did each of the 15 companies disclose?
The table below lists the headline disclosed numbers for each company as of the February to April 2026 collection window, with the channel it used.
| Company | Headline disclosed numbers (2025 to April 2026) | Primary disclosure channel |
|---|
| OpenAI | 900M weekly active users (Feb 2026); 50M paying subscribers; 92% of Fortune 500 deployed; more than 7M enterprise seats; $25B+ ARR (Feb 2026) | Company blog, funding announcements |
| Anthropic | About $30B annualized revenue (April 2026); 300,000+ business customers; 500+ customers spending $1M+ a year; 8 of the Fortune 10; $380B valuation (Feb 2026) | Company blog, funding announcements |
| xAI / Grok | 17.8% US chatbot share (Jan 2026); about 60M monthly active users; $230B xAI valuation (Jan 2026); about $500M ARR | Founder posts, Reuters reporting |
| Google DeepMind / Gemini | 750M monthly active users (per Google); 7M+ developers building on Gemini | Google I/O, company blog, investor calls |
| Microsoft AI / Copilot | 15M paid M365 Copilot seats; 33M active users; 35.8% workplace conversion rate | Quarterly earnings, Microsoft Build |
| Meta AI | 1B+ monthly active users across Meta platforms; Llama open-source adoption | Zuckerberg on quarterly earnings, company blog |
| Perplexity | 45M active users; 780M monthly queries; $200M+ ARR (Sept 2025); $20B valuation (late 2025) | Company blog, funding announcements |
| Mistral | European sovereign-AI positioning; open-weight model releases; funding announcements | Funding announcements, research papers |
| Cohere | Enterprise-only positioning; Command model releases; Fortune 500 customer disclosures | Enterprise partnership announcements |
| Character AI | Fell from third to ninth in the AI consumer category; acquired by Google in 2024 | Limited disclosure after the acquisition |
| Midjourney | Discord-first distribution; subscription-only model; no public revenue or user disclosure | Near-zero corporate communications |
| Runway | Gen-3 and Gen-4 video model releases; Hollywood creative-industry positioning | Product launches, creator testimonials |
| Hugging Face | Open-source model hub; developer-first positioning; funding rounds | Community posts, open-source releases |
| Stability AI | No systematic public disclosure of material usage or financial figures | Minimal |
| Inflection AI | No systematic public disclosure of material usage or financial figures | Minimal |
How are the four transparency tiers defined?
The transparency tiers group the 15 companies by how often and in how much detail each one discloses material information about users, revenue and enterprise customers.
| Tier | Companies | Disclosure pattern |
|---|
| Tier 1, high | OpenAI, Anthropic, Google DeepMind / Gemini, Microsoft AI / Copilot | Concrete usage, revenue and enterprise customer figures at least quarterly |
| Tier 2, moderate | Meta AI, xAI / Grok, Perplexity | Episodic figures tied to funding rounds, launches or competitive moments |
| Tier 3, low | Mistral, Cohere, Runway, Hugging Face | Rare disclosures tied to fundraising or acquisitions |
| Tier 4, minimal | Midjourney, Character AI, Stability AI, Inflection AI | Near-zero systematic disclosure of material usage or financial figures |
Which AI founders speak for their companies?
Five of the 15 companies are founder-dominant: Sam Altman at OpenAI, Elon Musk at xAI, Dario Amodei at Anthropic, Aravind Srinivas at Perplexity and Clem Delangue at Hugging Face. Dual-voice or corporate-dominant companies split the role. That group includes Demis Hassabis at Google DeepMind, Satya Nadella and Mustafa Suleyman at Microsoft, Mark Zuckerberg and Yann LeCun at Meta, Aidan Gomez at Cohere and Arthur Mensch at Mistral.
Silent, acquired or minimal voices include David Holz at Midjourney, Noam Shazeer after the Character AI acquisition, Emad Mostaque after his departure from Stability AI, and Cristobal Valenzuela at Runway.
In the Everything-PR Research Team's assessment, three of the five founder-dominant companies carry founder-level reputation risk that reaches the corporate brand. Buyers of those vendors take on the founder's public posture along with the product. EPR explores that dynamic in Who Controls AI Answers.
What should CIOs, CMOs and boards do with these findings?
CIOs, CMOs and boards should compare AI vendors only on figures the vendors have published, and treat a missing figure as a question to ask. Companies in the high-transparency tier can be benchmarked against one another with real data. Companies in the low and minimal tiers cannot. A vendor's communications posture is itself information about its scale, customer roster and durability.
Before a seven-figure AI contract, ask each shortlisted vendor for its latest active-user, seat and revenue figures in writing, with the date of each. If it declines, record that as a risk in the procurement file.
Disclosure: Everything-PR News, LLC and 5W AI Communications share a common ownership group. Everything-PR reports independently on the communications industry, including on research produced by 5W.
