Bot traffic used to sit in the analytics hygiene bucket. The newer problem is more commercial: non-human visits can enter retargeting pools, change how audiences price in auctions, and make open-web media look less efficient than closed platforms with logged-in identities.
PPC Land reported on September 8 that bots and AI agents now account for more than half of web requests, citing Cloudflare data and agency reports of retargeting pools filling with non-human traffic. GoFish clients were described as seeing bot traffic rise 80% year over year, with retargeting adjustments pushing CPMs up by an average of 20%.
Why the same signal creates two opposite decisions
For some brands, agentic traffic is attractive. If an assistant sends a ready-to-buy visitor, that session can convert better than the site average. For a media buyer, the same session can poison an audience if it is treated as a normal human visit and later used for retargeting.
The decision is therefore not whether bots are good or bad. The decision is whether each class of automated visit is allowed to enter marketing systems that spend money. Search crawlers, AI answer engines, shopping agents, uptime monitors and fraud traffic should not all receive the same treatment.
A practical audit
Start with four lists. First, map pages that feed paid audiences: cart, pricing, product, lead form, demo and high-intent content URLs. Second, identify known automated traffic by user agent, IP intelligence, data-center patterns, impossible session timing and request concentration. Third, compare retargeting audience growth with real downstream events such as qualified leads, orders and CRM progression. Fourth, separate traffic you want for discovery from traffic you do not want inside paid activation.
The control point is not only a bot-blocking rule. It is the handoff between site traffic, analytics, customer-data platforms and ad audiences. A visit can be allowed for indexing or product research while still being excluded from a remarketing list.
What marketers should change now
Do not judge the issue by total sessions. Look for audiences that grow while conversion quality falls, CPMs that rise after bot-filtering changes, and campaigns where frequency accumulates against visitors who never perform human downstream actions. If those patterns appear, the fix is a measurement contract: which automated traffic is visible, which is excluded, and which is quarantined until it proves value.
Closed platforms and retail media networks benefit when open-web audiences look unreliable. That does not mean every budget should move there. It means open-web buys now need traffic-quality evidence in the same conversation as reach, frequency and cost.
Sources
- PPC Land: Half of web requests are bots as agency CPMs climb 20%
- Cloudflare: content independence and AI traffic controls
