Google's OKF, ChatGPT Shopping at Scale, and the sameAs Signal Agents Trust Most
This week: Google published the Open Knowledge Format — the first structured spec for agent-readable business knowledge. ChatGPT's shopping agent crossed 800M weekly users. And the one Schema.org property that AI verification agents look for more than any other. Here's what moved, and what to fix.
This Week's Signals
Google Published the Open Knowledge Format (OKF v0.1)
Why it matters for your score
OKF is Google's spec for packaging your business knowledge as a directory of linked markdown files at /okf/ — designed for AI agents to navigate natively without relying on fragile web scraping. If your domain doesn't have /okf/index.md, you score 0 on this emerging check. Faro now scans for it as the third tier of Structured Data. This is currently a differentiator: fewer than 0.1% of sites have it. Building it now puts you ahead of the wave, not behind it. Faro's OKF Generator is in development to automate this.
ChatGPT Shopping Hits 800M Weekly Active Users
Why it matters for your score
ChatGPT's shopping and product recommendation features are now operating at consumer scale. When a user asks 'what's the best tool for X', ChatGPT pulls structured signals from your site — pricing clarity, schema markup, llms.txt content — not your design or branding. If your pricing page doesn't have machine-readable price numbers and plan names, you're invisible in those comparisons. The Pricing Clarity Auditor in Faro's dashboard checks exactly this.
Perplexity Expanded the Sonar API — Agents Can Now Query It Directly
Why it matters for your score
Perplexity's Sonar API is now embedded in dozens of enterprise software products, meaning AI agents built inside other tools can query your domain through Perplexity's index. Sites without llms.txt or structured entity data get lower confidence scores in those responses. If a procurement agent is evaluating vendors through Sonar and your domain is unstructured, you lose the comparison — before a human ever sees the results.
sameAs Is Now the Entity Verification Signal AI Agents Trust Most
Why it matters for your score
sameAs is a property inside your Organization or SoftwareApplication JSON-LD schema that links your website to Wikidata, LinkedIn, Crunchbase, or other authoritative sources. AI agents — ChatGPT, Perplexity, Claude — use sameAs to verify a business's identity across knowledge graphs. Without it, agents treat your business as unverified. With it, you're a confirmed entity. Faro's June scoring update added 9 points to sameAs in the Structured Data category — it's now the single highest-impact check you can add in under 20 minutes.
agents.json Emerges as a Proposed Standard for Agent-Callable Actions
Why it matters for your score
Similar to how robots.txt tells crawlers what to index, agents.json is a proposed standard for publishing what AI agents can actually do on your site — place orders, retrieve pricing, book demos. It's early-stage and not yet in Faro's audit, but it's moving fast. We're tracking it and will add it as a check when the spec stabilises. Watch this space.
Emerging Audit Signals
OKF Bundle Detection (/okf/index.md)
Faro now checks for the presence of /okf/index.md on your domain. This is Google's Open Knowledge Format — a linked markdown knowledge graph for AI agents. Currently worth 2 points in the Structured Data category. We expect this weight to increase as more AI systems start requiring it.
sameAs Entity Resolution (Tier 2 Structured Data)
A new 9-point check added in the June update. Faro looks for sameAs links inside your Organization JSON-LD pointing to Wikidata, LinkedIn, or Crunchbase. This is now one of the highest-weighted single checks in the entire audit. If your JSON-LD has no sameAs, you're leaving 9 points on the table.
Pricing Page Machine-Readability
The Pricing Clarity Auditor checks whether price numbers, plan names, and a comparison table are present and readable by agents — not just visually rendered in a way humans can see but bots cannot parse. JavaScript-rendered pricing tables that don't hydrate server-side frequently fail this check.
What to do this week
- 1
Add sameAs links to your Organization schema today — link to your Wikidata entry, LinkedIn company page, and Crunchbase profile. This is a 9-point score gain and takes under 20 minutes. Use Faro's AI Schema Creator to generate the updated JSON-LD.
- 2
Check whether GPTBot, ClaudeBot, and PerplexityBot are explicitly allowed in your robots.txt. Run Faro's robots.txt Analyzer — any of these blocked is an immediate disqualifier for AI recommendation.
- 3
Generate your llms.txt file if you don't have one. It's the fastest way to give AI systems direct context about who you are, what you do, and who you serve. Use Faro's llms.txt Generator in the dashboard.
- 4
Audit your pricing page for machine-readable price numbers. If your prices are rendered via JavaScript with no server-side content, procurement agents can't see them. Add static text pricing as a fallback minimum.
- 5
Run a fresh Faro scan this week — the June scoring update changed the Structured Data weights. Your current score likely looks different from two weeks ago, and you want to see the updated numbers before the next monitoring cycle.