2026-07-28 Final Spec Ships Today; Primary Research Pages Average 11.3 Citations
The 2026-07-28 MCP specification, described by maintainers as the largest revision in the protocol's history, is not backward compatible with older clients. Separately, a Growth Memo study found primary research pages averaged 11.3 citations per page versus 3.4 for all other pages — but only 2.7% of cited pages qualified as genuine primary research.
This Week's Signals
MCP 2026-07-28 Final Spec Ships Today: Stateless Core, 10K+ Servers, 97M Monthly SDK Downloads
Why it matters for your score
Servers built on the 2026-07-28 revision may not work with older clients, which means any business running an MCP server needs to check client compatibility now. The protocol now has more than 10,000 public servers in production, and OpenAI, Google, Microsoft, and AWS have all built MCP into their agent stacks. Deprecated features including Roots, Sampling, and Logging remain functional for at least 12 months under the new formal deprecation policy.
Growth Memo Study: Primary Research Pages Are 3.3x More Citation-Dense in LLMs, But Only 2.7% of Cited Pages Qualify
Why it matters for your score
Only 2.7% of cited pages — 8 of 301 — qualified as genuine primary research, yet they earned 8.4% of all citations (90 of 1,075). Original data does not automatically win LLM citations; the study found content must also be structured for AI extraction. Key citation drivers identified were entity-rich comparisons, named competitors, first-party data, and transparent methodology.
Contently / Anagram Data: AI Search Visits Up 42.8% YoY to 27.4B in Q1 2026; 87.4% of AI Referral Traffic From ChatGPT
Why it matters for your score
AI search visits grew 42.8% year-over-year, from 15.6 billion to 27.4 billion between Q1 2025 and Q1 2026. With 87.4% of AI referral traffic originating from ChatGPT, blocking OAI-SearchBot is the single decision most likely to remove a brand from AI-generated citations. Cloudflare customers who block crawlers classified as 'Training' will also block Googlebot, Applebot, and BingBot starting September 15, 2026, because those are classified as mixed-use crawlers.
What to do this week
- 1
Verify your structured data is present and valid (/tools/ai-schema).
- 2
Run a Faro scan on your own site (/tools/ai-readiness-scan) and note which checks fail.
- 3
Open your robots.txt and confirm the AI crawlers named in today's stories are allowed (/tools/robots-txt).