AI News/AI News

Traffic pool is shrinking, so track total visits and prepare for agents

Duane Forrester argues the AI versus Google traffic debate misses a falling total, and marketers should focus on offers that AI agents can read and complete.

Key points

  • Similarweb data for 10,000 domains per market shows U.S. organic visits down 13.2% over two years, with France down 24.4%, per Kevin Indig.
  • Paid search growth covered only about a third of the U.S. organic loss, and AI referrals make up roughly one visit in 200.
  • A McGill study of 100 hotel listings found agents converged on one listing that took 78.2% of bookings, so visible attributes beat placement.
  • Google's Universal Cart launched in the U.S. on Oct. 15; WebMCP remains an origin trial in Chrome.

Why it matters for marketers: Comparing AI referrals with Google organic hides the bigger issue: total inbound visits are falling, so teams should track their whole traffic pool and prepare offers agents can read.

Writing for Search Engine Journal, Duane Forrester says last month's debate over AI referrals versus Google organic missed the larger point. Studies from Marie Haynes, The SEO Works and Kevin Indig all showed AI sends far fewer visits than Google, and many readers concluded that generative engine optimization is a distraction. Forrester agrees Google still sends more, but says that ratio ignores the shrinking total both channels draw from.

He points to Indig's Similarweb analysis, where organic visits dropped over two years in both the U.S. and France, and publishers took the heaviest hit. Paid search growth offset only a portion of the loss, and email was the steadiest gainer in Indig's U.S. sample. A Bocconi preprint adds that ChatGPT sessions end in an outbound click far less often than Google queries do.

For marketers, the practical shift is measuring total inbound traffic across all sources, set against impressions for category queries, rather than watching a channel ratio. Forrester says the goal moves from the visit to the completed action, such as a booking or order an agent finalizes.

He cites McGill research suggesting agents pick offers based on readable attributes like price and reviews, not list position. A companion study found agents working for a platform chose sponsored listings far more often than agents working for the traveler. His checklist starts with the offer, structured reputation and policy data, complete feeds and consistent facts, then bot rules and commerce protocols.

Watch the infrastructure's maturity. Universal Cart is live in the U.S., while WebMCP is still a trial. Forrester also warns that some Cloudflare defaults may block agents without site owners knowing, so check those settings.

Sources

This brief is our own summary. Read the full reporting at the source.

You read about AI automation. We run it.

DocFluence, built by Z Media, writes, designs, films and publishes for your brand every day, in your voice, from one login.