Key takeaways
- GA4 removes known bots automatically using Google research and the IAB's International Spiders and Bots List, but you cannot switch that off or see how much it removed, and bots that pose as browsers get through.
- Bot traffic shows up as a pattern, not a single visit: one source or placement with near-zero engagement, a flat volume at night, one network or device profile, and orders that do not move when sessions jump.
- Shared IPs, VPNs and zero-second visits are weak evidence on their own. Google's own invalid-traffic page says zero-second visits and shared IPs often come from how analytics measures, not from fraud.
- Size the problem per placement or campaign, not per site. A cost per clean session and a CPA per placement tell you what to exclude; a site-wide bot percentage does not.
- Google Ads does not charge for invalid clicks it catches before billing and credits what it finds later; the Invalid clicks column and the Invalid Activity Credit Report show both.
Bot traffic detection means finding the visits in your analytics that no real person made, so they stop distorting your traffic, conversion rates and ad costs. GA4 already removes known bots on its own, but it cannot see bots that pose as an ordinary browser. Those show up as patterns you can find yourself: a source with near-zero engagement, a flat volume at night, one network or device profile, and sessions that rise while orders stay put. You confirm the pattern in your server logs and your order data, then fix it where it starts: the ad placement, the form or the firewall.
Bot traffic is any visit or request made by software rather than a person. Some bots are useful, such as search engine crawlers. Others scrape content, submit spam or click ads. In advertising, the umbrella term is invalid traffic: activity that does not represent a real, interested visitor and should not be counted or paid for.
What is bot traffic, and which bots should you worry about?
Bot traffic covers everything from Googlebot indexing your pages to scripts clicking your ads, and only part of it is a problem for marketers. Imperva’s 2025 Bad Bot Report estimates that automated traffic passed human traffic in 2024, at 51% of all web traffic, and that malicious bots alone made up 37%, up from 32% a year earlier. Most of it never appears in your analytics, because simple bots do not run the JavaScript your analytics tag needs.
The bots that do reach your reports fall into three groups:
- Declared crawlers. Search engines, SEO tools and uptime monitors say who they are in the user agent. Analytics tools filter most of them out.
- Disguised automation. Scrapers, price checkers and headless browsers that load your pages with a normal-looking user agent. They inflate sessions and page views and dilute conversion rates.
- Paid-media fraud. Bots, click farms and badly placed ads that generate clicks you pay for. Google Ads groups these under invalid clicks: accidental clicks, automated traffic and manual clicks meant to raise your costs or a publisher’s income.
The Media Rating Council, which audits ad measurement in the US, defines invalid traffic as activity that does not represent legitimate traffic that should be counted, and names non-human traffic such as spiders and bots as one of the main causes. Its guidelines separate general invalid traffic, which list-based filters catch, from sophisticated invalid traffic, which takes deeper analysis to find. That split is a useful way to think about your own checks: the list-based layer is already done for you, and the rest is pattern work.
Does Google Analytics filter bot traffic automatically?
Yes, partly. GA4 excludes known bot and spider traffic automatically, identified through Google’s own research and the International Spiders and Bots List maintained by the Interactive Advertising Bureau (IAB). Google’s help page is explicit about the limits: you cannot disable the exclusion, and you cannot see how much known bot traffic was removed.
Anything not on that list gets counted like a human visit. GA4’s data filters will not catch it either, because they are not bot filters. They exclude three things: developer traffic from debug mode, internal traffic from IP addresses you define, and traffic from hostnames you exclude. Once a filter is active its effect on incoming data is permanent and it does not apply to historical data, so test it first.
| Layer | What it removes | What it misses |
|---|---|---|
| GA4 known-bot exclusion | Declared crawlers and bots on the IAB list | Bots using a normal browser user agent |
| GA4 internal traffic filter | Your office or agency IP ranges | Anything outside the IPs you list |
| GA4 hostname filter | Hits sent to your measurement ID from domains you exclude | Bots that load your real pages |
| Google Ads invalid-click filtering | Invalid clicks before billing, with credits for later finds | Bots on other ad networks and organic or direct bots |
| Your own checks (this article) | Patterns in sources, engagement, timing and logs | Nothing is complete; you decide by evidence |
This is also one reason two analytics tools never report the same totals. Each one filters with its own list. The guide to GA4 data discrepancies shows how to compare tools on a page bots rarely reach, such as the thank-you page, to separate filtering differences from tracking bugs.
What are the signs of bot traffic in your reports?
Bot traffic rarely gives itself away in one visit. It shows up as a segment that behaves unlike your customers, so the first move is always a breakdown by source, campaign, placement and landing page. Look for a slice that ticks several of the boxes below at once.
| Sign | What it looks like | Innocent explanation to rule out |
|---|---|---|
| Sessions jump, orders do not | One source doubles in a week while sales from it stay flat | A new awareness campaign with a cold audience |
| Near-zero engagement | Average engagement time close to 0 seconds for one source or placement | Slow page, visitors leaving before the tag loads |
| Flat volume at night | The same number of sessions every hour, including 3 a.m. | A real audience in another time zone |
| One network or location | Sessions concentrated in one hosting provider or one city outside your market | VPN users; shared mobile-carrier IPs |
| One device profile | Identical screen resolution, browser version or an outdated user agent | A popular phone model in your market |
| Unknown hostname | Page views recorded on a hostname that is not your site | A staging site or a legitimate subdomain |
| Junk conversions | Sign-ups or leads with fake names, disposable emails or repeated phone numbers | A form without validation that real users also mistype |
| Conversion rate collapse | One campaign’s clicks rise and its conversion rate falls by half | Changed targeting or landing page |
Be careful with the right-hand column. Google’s own page on invalid traffic warns that zero-second visits and duplicate IPs often come from how analytics tools measure, not from invalid traffic, and that network address translation can put hundreds or thousands of devices behind one public IP. Many visitors in Iran browse through a VPN, so a foreign data center as the apparent location is weak evidence on its own for an Iranian site. Treat each sign as a reason to look closer, and a combination of three or four as a real suspect. Junk conversions are much easier to isolate when each lead event carries a form_id parameter and your CRM sends the later qualify_lead stage back as its own event, so fake sign-ups show up as leads that never qualify; the event tracking guide lists which parameters to send for a lead-gen site.
How do you detect bot traffic step by step?
Detection works best as a short, repeatable routine: find a suspicious segment in analytics, confirm it in a second data source, then trace it back to where it enters. Here is the order that wastes the least time.
- Break down by entry point. In your analytics, view sessions, engagement time and conversions by source / medium, campaign and landing page for the last 28 days. Consistent UTM parameters are what make this possible for paid traffic; without them, bots from five placements blur into one “referral” line.
- Flag segments with three or more signs from the table above. Write down the segment, the date range and the signs, so you can tell later whether a fix worked.
- Check by hour and by day. Real traffic follows your audience’s day. Automation often runs flat or in bursts at fixed times.
- Confirm in server logs. Your web server or CDN logs every request, including bots that never run your analytics tag. Look for the same IP ranges, user agents and timing as the suspect segment. If logs show far more requests than your analytics shows page views, most of that gap is bots your tag never saw.
- Verify crawlers that claim to be Google. Googlebot is often impersonated. Google’s method is a reverse DNS lookup on the IP, a check that the name ends in googlebot.com, google.com or googleusercontent.com, and a forward lookup that must return the same IP. Google also publishes its crawler IP ranges as JSON files.
- Compare against money. Match the suspect segment against orders or qualified leads in your backend. A segment that sends 30% of sessions and 2% of orders is your priority, whatever its cause.
- Trace it to the source. For paid traffic, go down to placement, publisher or keyword level. For organic or direct traffic, look at the landing pages and forms the bots hit.
The website tracking guide covers setting up the events and purchase data that make step 6 possible.
How do you measure click fraud on paid channels?
Click fraud is measured per campaign and placement: compare paid clicks with sessions that show real engagement, then compare cost per clean session and cost per order across placements. The ad platform’s own filters are the first layer. Google Ads says you are not charged for invalid clicks it detects before billing, and that invalid activity it finds after billing is credited as a billing adjustment. You can add the Invalid clicks column to your campaign table, and the Invalid Activity Credit Report breaks credits down by campaign and network for Search and Performance Max campaigns.
Two details from Google’s documentation matter when your own numbers disagree with the platform’s. Google Ads removes the invalid clicks it filters from its reports, but your analytics may still record those visits. And there is no industry standard for a normal invalid traffic rate, so a vendor quoting one as a benchmark is giving you their own number.
A worked example
Say, hypothetically, a store spends $2,000 on a native ad network over one month. The network bills 4,000 clicks at $0.50 each. The store’s analytics records 3,100 sessions tagged with that campaign, which is a normal gap: some people leave before the page loads. Of those sessions, 900 share four signs: about 0 seconds of engagement, one hosting network, the same screen resolution and steady volume between 2 and 5 a.m. The campaign produced 44 orders.
| Metric | All sessions | Excluding the 900 suspect sessions |
|---|---|---|
| Sessions | 3,100 | 2,200 |
| Conversion rate | 44 ÷ 3,100 = 1.42% | 44 ÷ 2,200 = 2.0% |
| Cost per session | $2,000 ÷ 3,100 = $0.65 | $2,000 ÷ 2,200 = $0.91 |
| Cost per order | $2,000 ÷ 44 = $45.45 | $45.45 |
Cost per order does not change, because the orders were real. What changes is the diagnosis. The campaign does not have a weak landing page converting at 1.42%; it has a landing page that converts real visitors at 2.0% and a bot problem. Broken down by placement, the picture gets sharper. Say placement C took $500 of the budget, produced 700 of the 900 suspect sessions and 2 orders: a cost per order of $250. The other placements spent $1,500 for 42 orders, or $35.71 each. Excluding placement C and moving its budget is the decision, and you would not see it from site-wide averages. The same logic carries into ROAS and every other budget call in the performance marketing metric chain: judge return per placement, on clean data.
How do you filter or block bot traffic once you find it?
Fix bot traffic where it enters, and keep your reporting honest rather than deleting history. Each source of bots has its own fix:
- Ad placements and publishers: exclude the placements, apps or sites that sent the suspect segment, and ask the network for a credit with your evidence. On Google Ads, use the Click Quality Form within the 60-day window Google allows for investigations.
- Forms and sign-ups: add server-side validation, rate limits and a challenge on suspicious submissions, so junk leads never become conversions.
- Scrapers and floods: use your host’s or CDN’s firewall, rate limiting or bot management. Robots.txt will not help here: Google’s documentation says it cannot enforce crawler behavior, and that only respectable crawlers obey it.
- Your own and your agency’s visits: add them to GA4’s internal traffic filter, test it, then activate it.
- Reports: for past periods you cannot clean, add an annotation and build a segment that excludes the suspect source. Report the clean figure next to the raw one, so nobody compares a cleaned month with a dirty one.
How does AdPix handle invalid traffic?
AdPix includes fraud monitoring from the Starter plan up, with invalid-click reports, an allowlist and blocklist export, and adds automatic fraud blocking and an evidence ledger for refund claims on Pro, so the checks above do not depend on one analyst remembering to run them each month. AdPix is a web analytics platform that replaces Google Analytics 4: realtime, acquisition, engagement, conversion and e-commerce reports, plus attribution, lift testing and marketing mix modelling, with no data sampling and a free plan.
A few points matter for bot work specifically. Reports are not sampled, so a small suspect segment such as one placement on one landing page is counted in full rather than estimated. The GA4 data sampling guide shows when GA4 explorations switch to estimates, which is exactly when a thin bot segment is hardest to trust. Acquisition and UTM campaign reports are on every plan, including Free, which covers steps 1 to 3 of the routine. AdPix runs alongside Google Analytics without interfering, so you can compare both tools on the same pages while you investigate. For Iranian sites, keeping a second tool next to GA4 is also the low-risk setup recommended in the guide to Google Analytics risks in Iran. Starter costs 499,000 toman a month and adds audience segments, custom reports and API access next to fraud monitoring; see AdPix pricing for the full plan list. How AdPix scores each paid click and traffic slice, and what each verdict asks you to do, is covered in the AdPix invalid traffic detection guide. The AdPix features overview explains how invalid traffic detection fits with lift tests and MMM, and the Google Analytics alternative guide compares GA4, Matomo and AdPix on the rest. If you would rather keep raw logs and analytics on your own servers, the Matomo vs Google Analytics comparison shows what self-hosting costs in servers and upkeep.
Start with one paid campaign this week: break it down by placement, mark every segment with three or more signs, and compare its cost per order with the rest before you renew the budget.
CPM · CPC · CTR calculator
Cost and click metrics for any campaign. The free plan covers 100,000 page views a month for each website, no credit card. Setup takes under five minutes.
Frequently asked questions
How do you detect bot traffic in Google Analytics?
Break traffic down by source, campaign and landing page and look for a segment with near-zero engagement time, a flat volume at night, one city or network, and no orders. Then confirm it outside GA4: check your server logs for the same pattern and compare the segment against orders in your shop backend.
Does GA4 filter bot traffic automatically?
Yes. GA4 excludes known bots and spiders using Google research and the International Spiders and Bots List maintained by the IAB. You cannot turn this off or see how much was excluded, and bots that imitate a normal browser are not on that list, so some still reach your reports.
What percentage of website traffic is bots?
Imperva's 2025 Bad Bot Report estimates that automated traffic made up 51% of all web traffic in 2024 and that malicious bots made up 37%. Your own share depends on your site and channels, and much of it never reaches your analytics because it does not run JavaScript.
Is click fraud the same as bot traffic?
Click fraud is the paid-media part of the problem: clicks on ads that do not come from a real, interested person, whether from bots, click farms or accidental placements. Bot traffic is broader and also includes scrapers, crawlers and spam that never touch an ad.
Does blocking bots in robots.txt stop bot traffic?
No. Google's documentation says robots.txt cannot enforce crawler behavior and that only respectable crawlers obey it. It is useful to manage good crawlers, while bad bots need a firewall, rate limits or bot management at your host or CDN.
Sources and methodology
GA4, Google Ads and Google Search behavior come from the Google help and Search Central pages in the sources, read on 7 November 2026. The bot-share figures are Imperva's own estimates for 2024. The campaign in the worked example and all of its numbers are hypothetical; the ratios are simple calculations made in this article. AdPix facts come from adpix.io.
- Google Analytics Help: [GA4] Known bot-traffic exclusion
- Google Analytics Help: [GA4] Data filters
- Google Ads Help: Invalid clicks: definition
- Google Ads Help: About invalid traffic
- Google Ads Help: About the Invalid Activity Credit Report
- Imperva: 2025 Imperva Bad Bot Report
- Google Search Central: Verifying Googlebot and other Google crawlers
- Google Search Central: Introduction to robots.txt
- Media Rating Council: Invalid Traffic Guidelines update, public comment announcement (2020)
This article was drafted with AI assistance and reviewed by the AdPix team before publishing. Editorial policy