Meta AI Bots Hit Publishers With 9 Billion Requests

Meta’s AI bots generated 9 billion requests in Q2 2026, but DataDome analysis says the activity returned zero traffic to publishers. The bot-management and agent-control company also found that agentic traffic surged 45% during the quarter, while Meta AI bots increased 163% from Q1.

Every crawl consumes bandwidth, server resources, logging capacity and CDN transactions. Unlike traditional search crawlers, which can send visitors back to a site, AI crawlers may extract content without producing measurable referrals.

Meta’s crawler growth and ChatGPT referrals

Meta AI is typically associated with the chatbot inside Facebook, but the rapid growth of its Meta-WebIndexer bot suggests a broader effort to index websites, similar to the role Google’s crawlers have played for decades. DataDome said Meta now dominates AI traffic on its network.

ChatGPT crawlers moved in the opposite direction: their traffic declined, while ChatGPT accounted for 88% of AI referrals. The report suggests OpenAI’s systems may need fewer crawls because they can answer questions using information already collected about websites.

“Q2 showed us that the ground is shifting faster than most organizations realize. Meta now dominates AI traffic on our network, MCP traffic has emerged as a real signal, and ChatGPT is driving more referral value with fewer crawls.”

DataDome also identified growing activity associated with the Model Context Protocol (MCP), which connects AI agents to external tools and differs from conventional crawler traffic. Its emergence gives organizations another category to monitor—and another potential part of their attack surface.

Recommended reading

Dimon warns Claude Mythos could become a cyber weapon

Setting policies for AI agents

The report’s cybersecurity implications extend beyond ransomware, malware and phishing. Websites may need different access policies for different agents: full crawl access for Google, retrieval access for ChatGPT, and rate limits for Meta AI if its traffic produces little value. Unknown agents could face additional verification or be blocked entirely.

The central distinction is whether an agent behaves like a user, retrieves information or scrapes content. DataDome said organizations that build policies around those differences are gaining an advantage as agent-trust adoption accelerates.