Agentic trust platform emerges from stealth as AI reshapes buying decisions. Its first State-of-the-Category study finds 47% of skincare brands lack evidence supporting product claims.
SAN FRANCISCO--(BUSINESS WIRE)--M11 Labs emerged from stealth today, an agentic commerce lab that studies how AI is reshaping online buying and builds the tools to make it work. Alongside its debut, the company unveiled its M11 agentic trust platform and OpenMarket, a live agentic marketplace where brand agents compete for the recommendation and every claim is checked in real time, spanning millions of merchants and billions of products.
M11 Labs’ inaugural State-of-the-Category report, The Trust Gap in Skincare, found that 47% of skincare brands lack independently verifiable evidence supporting the claims their products are sold on. The report will be released Sept. 16.
The Shift to AI-Mediated Commerce
The way products are discovered and evaluated is changing. Search engines ranked pages. Social platforms rewarded engagement. AI agents evaluate evidence.
As consumers increasingly rely on AI to research, compare and select products, brands face a new challenge: ensuring product claims can be verified by the systems influencing purchasing decisions.
Unlike human shoppers, AI agents do not simply read claims. They look for supporting evidence, assess source credibility and determine whether claims can be independently verified. When evidence is missing, inaccessible or difficult to validate, an agent may discount the claim, lower its confidence in a product or move on to an alternative that provides clearer proof.
“Brands spent the last decade optimizing for algorithms that rewarded attention,” said Ankur Modi, founder and CEO of M11 Labs. “Systems that reward proof will shape agentic commerce. If an AI agent can't verify a claim, the brand may never make the shortlist.”
For brands, the challenge is increasingly commercial rather than regulatory. A claim supported by evidence that an AI system cannot find, interpret or validate may have little influence on recommendations, rankings or purchasing decisions. The result is a growing trust discrepancy between what brands claim and what AI systems can verify.
Measuring the Trust Divide
M11 Labs analyzed over 600 skincare brands and more than 15,000 evidence captures in June 2026. The striking result: Nearly half of brands lack independently verifiable evidence behind the claims used to market their products.
The findings in The Trust Gap in Skincare report indicate the challenge is typically not false claims but unverified ones. Among the product claims reviewed, 58% referenced or implied supporting research that was not publicly available. Of the evidence links associated with judged claims, only one in 20 directly confirmed the claim.
“The problem is often not that evidence doesn't exist,” said Ankur. “It's that the evidence isn't accessible, connected to the claim or readable by the systems increasingly informing purchase decisions. We have generated tens of thousands of findings across thousands of brands, built thousands of competitor profiles and performed well over 20,000 regulatory exposure scans against enforcement bodies spanning multiple jurisdictions.”
How the M11 Platform Helps Brands
When an agent reads a claim and finds no reachable evidence, it moves on, and no error message or lost-sale report ever tells the brand it happened. M11 agents surface that silent skip as a signal with an owner and a fix, turning an invisible loss into something a team can act on.
The M11 platform agents work through four questions a shopping agent asks of every brand.
- Accuracy. Can an AI trust what you say? The agents fact-check each claim and bind it to its evidence, or flag where the evidence is missing. Most flagged claims are not lies. They are true statements with nothing to back them, and to an agent, the two look identical.
- Completion. Is the product record comprehensive? A spec missing load capacity or a formula without an ingredient list is a constraint an agent can't clear, so it skips the product without a human ever seeing it. The agents build the record back from the maker's own data.
- Presence. Does an agent find you at all? The agents ask what real buyers ask, query every AI surface, and publish the data a machine needs to rank the brand.
- Competition. What is challenging your product and brand? M11 agents track the opposition and market shifts, assess what each means, and flag what matters.
All four feed one number, the M11 Commercial Score, which shows where a brand stands and what needs attention. Behind it, M11 Pulse shows the score, M11 Graph connects each claim to its evidence, M11 Proof builds the evidence, and Command manages the signals through to deployment. Each agent goes past finding a problem to drafting the fix. The agents propose the change, the brand approves it, and the agents deploy it.
Often the proof is already in the brand's files. The largest single category of flagged claims is a real study the brand never linked where it made the claim, so the cheapest fix is connecting evidence a brand already owns and deploying it as structured data a machine can read. When the existing trail won't hold, M11 Proof generates independent evidence from sources a machine will accept, including clinical databases and certification registries. Much of that proof lives where no scraper or foundation model can reach it, in ERP and lab reports, compliance records and certification registries, which is why a general-purpose chatbot or a single sales platform can't do this work.
M11 applies the same standard to every brand, customer or not. It does not test products, certify performance or audit manufacturers. The question it answers is narrower: Can the evidence behind a claim be independently verified?
“Everything the platform generates must be independently verifiable,” Modi said. "We don't gate it, don't censor it. And we don't change the result because a brand is a customer. A score means something to an agent precisely because it can't be bought."
Inside OpenMarket
OpenMarket is M11 Labs' agentic marketplace and its research engine. Shoppers say what they want, brand agents compete for the recommendation, and every claim an agent makes is checked in real time. It runs on the Universal Commerce Protocol, the agentic-commerce standard co-developed by Shopify and Google, with support from Amazon, Meta, Microsoft, Salesforce, Stripe, Etsy, Target and Wayfair. Because agents read live catalog data through the protocol, what a shopper sees in OpenMarket is what a brand actually publishes, updated as prices and offers change.
OpenMarket spans the open marketplaces where most commerce lives, including Shopify, reaching more than 3 million merchants and billions of products. The same scans that let agents compete also feed M11 Labs' research, including the skincare report, giving the company a continuously refreshed read on how product claims hold up to machine scrutiny across the open web.
Not Only Skin Deep
Skincare is the first category assessed through the M11 Labs Verifiability Index, a framework designed to measure how well product claims stand up to independent machine review.
The company plans to expand the index into additional consumer categories where AI-assisted purchasing is becoming more common, and product claims increasingly influence buying decisions.
About M11 Labs
M11 Labs is an agentic commerce lab that studies and builds the frontier use cases of AI-driven commerce and works alongside the industry to build the standards and tools for a market where agents do more of the buying. Its commercial program centers on the M11 agentic trust platform, which helps brands keep their claims accurate and verifiable and detect threats from competitors and the market for AI agents, marketplaces and regulators. M11 Labs was founded by repeat operators and researchers from Meta, Amazon and leading AI labs. The company is based in London and San Francisco. For more information, visit m11.ai.
Contacts
Media Contact
Sam Brancato
PRforM11@bospar.com
