Bot Traffic and Ad Spend: Why Your ROAS Suddenly Dropped
When ROAS drops without a creative or targeting change, check the traffic before the ads. Bots inflate clicks and sessions, drag down engagement, and feed the platform conversion data from visitors who were never going to buy. Compare the server logs against analytics, and the gap is usually the answer.
The drop with no creative cause
ROAS falls off a cliff on a Tuesday. Nothing changed in the ads: same creative, same targeting, same budget. The instinct is to rewrite the ad or kill the campaign, and that is usually wrong. Check the traffic first, because the ads did not get worse. The traffic did. Bot traffic costs money three ways: it loads the server, it skews the data, and some of it commits fraud, clicking ads and testing stolen passwords. A few thousand unlabelled bot hits can drop engagement rate and average engagement time so a healthy page looks broken, inflate pageviews on a landing page so a test looks like it worked, and fill the referral report with domains nobody has heard of. The platform then optimizes against that poisoned history, finding more of the visitors who look like the bots, because the bots are the ones converting in the data it sees.
The scale is the context. Automated traffic was 51 percent of all web traffic in 2024 and past 53 percent in 2025, per Imperva's Bad Bot Report, and bad bots alone were 37 percent. GA4 shows much less of it, because GA4 filters the bots on the IAB's known list automatically and many bots never run the tracking script at all. That is why the analytics look only mildly wrong while the ROAS is badly wrong: the bots that never run JavaScript never appear in GA4, and the ones that do are the polite minority. The server logs count everything. GA4 counts browsers that ran the tag. When the logs show ten times the requests GA4 shows, that is normal, and the gap is where the bots live.
Reading the drop, and fixing it
The fingerprints are consistent. A sudden spike in traffic from countries the store does not serve. Sessions with near-zero engagement time, single-page visits, and no second pageview. Referrals from unknown domains, and a Realtime report showing the spike live, in a country and on a page that do not match the campaign. Cross-check with Search Console: if GSC shows nothing and GA4 shows a flood, something crawled the store directly, not through search. Then open the server logs and read the user agents: named crawlers like GPTBot and PerplexityBot are the AI traffic worth tracking, and generic strings like python-requests, curl, or Go-http-client almost never matter as shoppers.
The fix is layered, and none of it is in the ad account. Filter the known bots out of analytics and exclude the countries the store does not serve at the reporting layer, without touching collection, so nothing is lost permanently. Block the abusive patterns at the network level, where the bots can be stopped before they load the server or click the ad. And give the platform clean conversion data, because an optimizer fed bot conversions will find more bot-like visitors. The ROAS usually recovers in days, not because the ads improved, but because the traffic stopped lying. The lesson is permanent: when the number moves and the creative did not, the traffic moved first. The classification half is GIVT versus SIVT and sneaker drops being a bot problem first.
Frequently asked questions
Why did my ROAS drop if I did not change the ads?
Because the traffic changed. Bots inflate clicks and sessions, drag down engagement, and feed the platform conversion data from visitors who never buy. The platform then optimizes toward more of them.
How much web traffic is bots?
Automated traffic was 51 percent of all web traffic in 2024 and past 53 percent in 2025, with bad bots alone at 37 percent, per Imperva's 2025 Bad Bot Report. GA4 shows much less, because it filters known bots and many bots never run the tag.
How do I spot bot traffic in GA4?
Look for sudden spikes with very low engagement time, single-page sessions, traffic from countries you do not serve, and referrals from unknown domains. Compare with the server logs: if the logs show ten times the requests GA4 shows, the gap is the bots.
Does bot traffic hurt SEO?
Not directly through rankings, but it slows the server, fills analytics with junk, skews the engagement numbers used for decisions, and consumes crawl budget, which limits how much of the site gets indexed.