Faro AI Signals·July 31, 2026

Linux Foundation Takes MCP Helm as llms.txt Reality Check Arrives

The MCP 2026-07-28 final spec is now live under the Linux Foundation's Agentic AI Foundation, reshaping how agents connect to business data. Simultaneously, fresh data from a 137,000-domain study shows that the most popular AI visibility tactic — publishing an llms.txt file — is reaching almost no AI crawlers at all.

ShareShare

This Week's Signals

1

Final MCP Spec Moves Under Linux Foundation, Deprecates Three Core Features

Source: Model Context Protocol Blog (official) ↗

Why it matters for your score

The 2026-07-28 spec is now the standard all four Tier 1 SDKs ship against, meaning any business exposing tools or data through MCP needs to audit against the new spec immediately. Roots, Sampling, Logging, and legacy HTTP+SSE transport are all deprecated with a 12-month offramp — teams running older integrations have a countdown clock. Governance moving to the Linux Foundation's AAIF signals that MCP is infrastructure, not a vendor feature, which raises the stakes for getting your implementation right.

2

97% of llms.txt Files Got Zero AI Crawler Visits in May 2026, Ahrefs Data Shows

Source: DevCommX (citing Ahrefs May 2026 bot analytics study) ↗

Why it matters for your score

Ahrefs analyzed 137,210 domains and found that 97% of valid llms.txt files received no requests at all during May 2026. Of the requests that did arrive, roughly 77% came from non-AI sources — SEO audit tools alone accounted for 21.7% while AI retrieval bots accounted for about 1.1%. No major provider has publicly committed to acting on llms.txt in production inference, so publishing the file alone is not a reliable path to AI visibility right now.

3

SaaS Audit: More Than Half Have an llms.txt Path, Under a Quarter Match the Spec

Source: Arobis AI ↗

Why it matters for your score

A July 2026 audit of 30 well-known SaaS companies found that 57.7% had something at /llms.txt, but only 23.1% — six companies — had a file that actually matched the spec with a correct H1, summary, and organized sections. HubSpot's strength in AI recommendations traced back to content depth and third-party citations, not the file itself. The study concludes that llms.txt is not a substitute for building genuine AI recommendation authority through content.

What to do this week

  1. 1

    Check your MCP integration against the 2026-07-28 final spec, specifically whether you rely on Roots, Sampling, Logging, or legacy HTTP+SSE transport — all are now deprecated with a 12-month offramp. Use Faro's /tools/ai-readiness-scan to flag outdated connection patterns before the clock runs out.

  2. 2

    Audit your llms.txt file against the actual spec requirements — correct H1, summary section, and organized content sections — since the Arobis data shows more than half of published files do not meet the standard. Use Faro's /tools/llms-txt to generate or validate a spec-compliant file.

  3. 3

    Shift AI visibility effort toward content depth and third-party citations rather than relying on llms.txt alone, given that the Ahrefs study found AI retrieval bots account for about 1.1% of llms.txt requests. Use Faro's /tools/competitor-intelligence to identify where competitors are earning citations that you are not.

Sources

← All editionsScan your site now →