Meta’s web bots made 9 billion Q2 requests: almost no traffic sent back

DataDome’s Q2 2026 report says Meta Platforms’ automated agents have become the biggest source of chatbot- and indexing-related crawling across its network. If you run a publisher site, this one’s worth reading: DataDome says it handled 17.7 billion automated requests in Q2, up 45% from the previous quarter, with Meta-ExternalAgent and Meta-WebIndexer driving much of that jump. Meta-WebIndexer by itself was up 163%.

ChatGPT Download

For publishers, that adds up to billions of requests chewing through bandwidth, server capacity, logs, and CDN resources, with almost no referral traffic coming back in return, DataDome says. Traditional search crawling usually sends something back. This doesn’t.

DataDome also says requests tied to OpenAI and ChatGPT fell 6% in Q2. Even so, ChatGPT still accounted for 88% of all chatbot referral traffic, which means publishers are getting more actual visitors from fewer bot requests.

Meta-WebIndexer’s growth may point to something bigger, too. DataDome says the increase suggests Meta could be building out a broader web index.

And the larger fight is getting pricier for publishers. According to DataDome, machine-generated search summaries can reduce publisher traffic by 20% to 60% and may be tied to about $2 billion a year in lost ad revenue. On top of that, Model Context Protocol, or MCP, traffic brings its own operational and security problems because it connects agents to outside tools and widens a publisher’s attack surface.

Publishers are pushing back in a few different ways: selective crawler access, rate limits, outright blocks, lawsuits, licensing deals, and pay-per-crawl experiments. The full report is in DataDome’s Q2 2026 findings.