Consider the last time ChatGPT recommended a project management tool to you, or Claude steered you toward a particular CRM vendor. Where did that suggestion originate? For the vast majority of brands, the unsettling answer is: somewhere in the algorithmic ether, unverified and largely beyond their reach.
Apurva Luty thinks about this problem constantly. The former product leader at Microsoft, Meta, and Discord launched Seattle-based Optimly with a premise that sounds almost obvious once you hear it: according to Luty, if AI agents are becoming gatekeepers to consumer attention, brands need a way to control what those agents say about them. According to the 'State of AI Brand Representation 2026', published in June 2026, roughly 80% of brands are materially misrepresented by AI models. Not just misunderstood—wrong.
The distortions aren't trivial. Incorrect pricing, outdated product specs, even cases of mistaken identity that border on absurd. In one particularly telling example, Hal9—an AI development platform—found itself repeatedly conflated with HAL 9000, the homicidal computer from 2001: A Space Odyssey. Not exactly the association you want when pitching enterprise clients.
According to Optimly, this isn't just a brand image issue. It's infrastructure. As conversational agents begin handling shopping workflows end-to-end, being invisible or factually wrong in their training data carries commercial consequences. A June 2026 study suggests behavioral lift occurs following AI name-drops, though specific methodology details or direct validation are not provided: when a model name-drops your product, users act on it.
The Architecture of Trust
At the core of Optimly's offering is something the company calls BrandVault, a two-tiered system designed to give brands write-access to the data AI models consume. The outer layer, the AI Brand Index, covers more than 60,000 brands across three dozen categories—a searchable registry that serves as the first place models look when a user asks about, say, collaboration software or sneaker brands.
The inner layer is where things get interesting. Through BrandVault's paid tiers, companies can verify ownership (via email, DNS records, or Shopify integration), then publish what Optimly terms "canonical facts"—structured data on identity, products, pricing, positioning, policies. This verified information flows directly to ChatGPT, Claude, Gemini, and Perplexity, becoming the authoritative record those models reference.
It's a system that in some ways mirrors what Google Business Profile did for local search, though with higher stakes. Get your hours wrong on Google, and a customer shows up to a locked door. Get your pricing wrong in an AI agent's memory, and you might never know the deal was lost.
Pricing reflects the tiered ambition. Basic index inclusion is free. At $199 monthly, Signal offers monitoring and basic corrections. Signal Pro, at $799 per month, includes what the company calls "Fix Cards"—weekly prioritized action items mapped to specific misrepresentations detected across models. Enterprise packages, priced individually, handle multi-brand operations. There's also an AI Press Room in beta, designed to push announcements directly into AI systems' memory—a concept that feels either prescient or premature, depending on how quickly agentic commerce actually materializes.
For the developer-minded, Optimly provides a Scout integration for Claude via the Model Context Protocol, alongside public API endpoints for brand profiles, buyer simulation, and intent-based search. The platform speaks machine: llms.txt files, an AI Agent Manifest, an OpenAPI spec bearing the formal title "AI Brand Trust Registry API."
Numbers, With Caveats
Measuring early traction in a space this nascent gets messy. The discrepancies in AI crawler request data—one page references roughly 100,000 AI crawler requests per week; another cites 34,877; a social media post from Forum Ventures mentioned 11,000—are due to different measurement windows and data subsets, according to sources. Perhaps; or perhaps the metrics are still stabilizing as the product finds its footing.
The Hal9 case study offers clearer lines. After implementing BrandVault, the company saw a 16x increase in AI-referred traffic over 60 days. Its Brand Authority Index score—Optimly's proprietary metric—climbed from 2.0 to 8.1. ChatGPT's accuracy in representing the company improved by 217%, and those identity-level errors (the HAL 9000 confusion) dropped to zero.
Optimly's own growth story leans on similarly striking figures: a claimed 64x increase in Google impressions over 30 days with no SEO budget, and an "AI fetches per Google click" ratio of 119:1. The implication being that for every human visitor arriving via Google, over a hundred AI agents were already there, quietly gathering information.
Brand pages in the index carry rolling refresh timestamps; recent profiles show analysis dates from earlier this summer. It's a living dataset, at least in theory.
The Broader Landscape
Optimly isn't operating in a vacuum. Profound is focused on monitoring AI answer engines and tracking sentiment. Experian announced something called Agent Trust back in April, though positioned more around identity verification for AI agents themselves than brand representation. Fime launched FACT, a trust layer for agentic commerce, a couple months back. Azoma and Firmly Connect are building merchant-side standards for how brands expose data to agents.
What differentiates Optimly, according to Luty, is write-path control. Monitoring tools tell you what AI says about you. BrandVault purports to let you correct it at the source, publishing directly into the models' knowledge layer rather than hoping they eventually crawl updated information.
GeekWire took note of the approach in a profile this summer, observing that "the chatbots are listening" when brands use BrandVault to essentially teach models about themselves. The company also earned a spot in the Washington Technology Industry Association's 13th Founder Cohort Accelerator late last year.
The Bet
Optimly's recent funding round details were mentioned in a social media post by Forum Ventures, suggesting the company closed a pre-seed round within the past month or so, with backing from AI House, Mighty Capital, angels from Flywheel Conference, and Forum Ventures as the lead investor, though no formal press release confirms it. This is typical for rounds of this size—announcements often follow once the company has something concrete to show for the capital.
The team remains small, between two and ten employees, all based in Seattle. The exact number depends on how you count contractors and part-timers, a haziness common at this stage.
The underlying hypothesis is straightforward, even if the execution isn't: AI agents will consolidate discovery power much like Google once did for search. Brands that invest early in verified presence within these systems will capture disproportionate recommendation share. Academic work exploring AI recommendation concentration across models, published around the same time as Optimly's research, hints that the pattern may already be forming.
Whether 60,000 verified profiles constitutes critical mass or merely a promising start depends entirely on how fast agentic commerce transitions from experimental edge case to default behavior. The index continues growing—brand pages analyzed daily, API endpoints live and documented, the Claude integration operational.
For now, it's infrastructure in search of an inevitable future. Or at least a future that seems inevitable if you're Apurva Luty, watching AI agents become the new middlemen in every consumer decision, and wondering who gets to write the script they're reading from.
