Bot traffic inflating your numbers? How to filter it out

By Ranjit Bhinge, founder of Blur Studio · Updated · 9 min read

Short answer: Bot traffic is visits to your site by software instead of people. Google Analytics 4, Google's free website analytics tool, automatically removes known bots and spiders, and you cannot turn that off or see how many it removed. What slips through is unknown bots, spam that sends fake events straight to your account, and your own team's visits. You fix it by spotting the pattern first, adding an internal traffic filter and a hostname filter, then blocking the worst offenders at your host or CDN (a service in front of your site).

Ranjit Bhinge

You think bots or spam are inflating your visitor numbers, and you want to find out how much, filter it, and trust your reports again.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

Start by checking whether the problem is real. A traffic jump with no matching jump in calls, form fills or sales is the classic sign. Look at three reports before you change anything: traffic by country, traffic by source, and engagement time. If a spike comes from one country you do not serve, one odd source, or sessions that all last zero or one second, you probably have bots or spam. If numbers are high but the pattern looks like normal customers, you probably do not.

This page covers what Google already filters, how to spot the rest, the exact filters worth setting in GA4, when to go further at the server or CDN, and how to show a client or boss a clean before and after. For the other direction, where tools undercount real people, see the page on who is visiting your website. If you are not sure whether AI crawlers are reading your pages, check that first.

This page may contain affiliate links from which Blur Studio earns commissions.

What Google already filters, and what it does not

Google's help page says that in Google Analytics properties, traffic from known bots and spiders is automatically excluded. It also says that at this time you cannot disable known bot exclusion or see how much known bot traffic was excluded. The list comes from Google's own research and the International Spiders and Bots List kept by the IAB (the Interactive Advertising Bureau, an advertising trade group).

That means three things for you:

  • You cannot measure what GA4 already removed, so you cannot check Google's work.
  • Unknown or new bots are not on the list. A bot that pretends to be a normal browser looks like a person.
  • Do not say "GA4 filters all bots" to a client. Google only claims known ones.

Also remember that bots mostly do not run the tracking script on your pages. Many simple crawlers read the raw page and leave, so they never show up in GA4 at all. The numbers that inflate GA4 are usually bots that run a full browser, spam sent straight to your measurement ID, and humans you did not mean to count, such as your staff. The server log, not GA4, is where plain crawlers appear.

Step 1: spot the pattern

Open GA4 and check these in order. Use the last 30 days, then compare with the same days the month before.

CheckWhere to lookBot or spam sign
Sudden spikeReports, Acquisition, Traffic acquisition, by dayOne or two days far above normal with no campaign or press
Wrong countryReports, Demographics, countryThousands of sessions from a country you do not serve
Source or hostnameTraffic acquisition, first user source; hostname as a secondary dimensionOdd referral names, or a hostname that is not your domain
EngagementAverage engagement time and engaged sessionsNear zero seconds and one page per session across a large block
Direct trafficChannel group DirectDirect traffic climbing while organic and paid stay flat
ConversionsKey eventsTraffic up, key events flat or down

One caution on thresholds: GA4 can hide small rows in some reports. Google says system data thresholds cannot be changed, and a data quality icon appears when thresholding is applied. So a very small spam source may not show up at all in some views.

Step 2: set the GA4 filters that do work

Internal traffic filter

Your own team can inflate small-site numbers. Per Google's help page, the steps are: in Admin, open Data streams, choose the web stream, open Configure tag settings, Show more, then Define internal traffic. Name the rule, keep the traffic_type value "internal" (the default), and enter your office IP address or a range. Then go to Admin, Data filters, create a filter of type Internal Traffic, and set it to Exclude.

  • A filter has three states: Testing, Active and Inactive. Testing lets you check results, which Google says take 24 to 36 hours to appear, before you commit.
  • Google says an Active filter is permanent: it changes data from that point on and cannot be undone for past data. Use Testing first.
  • A property can have up to 10 data filters, and you need the Editor role.
  • This works for web data only. Remote staff on changing home addresses will not be covered by one IP rule.

Hostname filter

Spam sent straight to your measurement ID often carries a hostname that is not yours. Google's Data filters page lists a hostname filter that excludes data from other domains. Check the page for the current steps, then apply it in Testing state first and compare. Do not apply a hostname filter if you collect data on several of your own domains or on a payment provider's domain that you want to see, unless you list all of them.

What cannot be fixed afterwards

Filters apply going forward. If the last six months are already polluted, you can annotate them and compare periods, but you cannot scrub the history. Say that plainly in client reports.

Step 3: block the worst at the host or CDN

Filtering in analytics hides bots from reports, but they still hit your server. If they slow the site or fill your forms, block them earlier.

  • Cloudflare, a service that sits in front of many sites to speed up and protect them, gives each request a bot score from 1 (certain automated) to 99 (certain human). Per Cloudflare's docs, scores 2 to 29 are likely automated and 30 to 99 likely human. Pro plan or higher can see bot traffic groups in Security, Bots. Granular scores need Enterprise with Bot Management, and the page does not state what the Free plan shows.
  • Cloudflare's Bot Fight Mode is free and challenges traffic that matches known bot patterns. Its docs say it cannot be bypassed with WAF custom rules or Page Rules (a WAF is a web application firewall, a set of rules that block unwanted requests), so a service you rely on, such as an uptime checker, could be challenged. Check the page for which plans offer the finer Super Bot Fight Mode.
  • Form spam is a separate problem that also distorts lead counts. See how to stop spam on your contact form.
  • Scrapers that copy your content are covered in how to stop bots and scrapers.

Cloudflare also describes verified bots: bots that identify themselves accurately and follow rules such as robots.txt and reasonable request rates. Blocking all bots would also block search engines. Block by behavior or by list, not by "bot" as a word.

Step 4: compare to a count the bots cannot fake

The best check is a number outside GA4 (a count the bots cannot fake). Pick one of these:

  • Form leads or calls you actually received in the month. Real leads do not inflate with bots.
  • Requests in your server or CDN log, filtered to exclude known bot names.
  • A second analytics tool, with a different method.

Be careful on the other side. Ad blockers and privacy settings can hide real visitors from GA4. In one vendor test by Plausible (a privacy-focused analytics tool), over 3 days in August 2021, on a personal blog with a post trending on Hacker News and Reddit (two sites where people share and discuss links), Plausible counted 50,947 visitors where Google Analytics counted 21,054, so Google missed 58.67 percent. That was an extreme, tech-heavy audience and a vendor study, so do not use it as a general rate. The practical lesson is that GA4 may be too high because of bots and too low because of blockers at the same time. Compare leads to key events over a month to estimate your own gap.

If you host on Ploy, an AI website builder and host, its analytics docs say the Bots card counts AI crawlers separately from agent visits (an AI assistant opening a page for a person), and its changelog for 9 November 2026 says traffic is now split into people, routine crawling and agent visits. The docs do not describe filtering rules for the human totals, and they do not state a data retention period, so ask Ploy how your numbers are cleaned before you quote them. Analytics is included on every Ploy plan per Ploy, which as of November 2026 are Free at $0, Starter at $50 a month, Pro at $300 a month and Enterprise by quote.

A worked example with made-up numbers

A dental practice sees 6,400 GA4 sessions in October, up from 3,100 in September, with 38 booking-form key events in both months. The owner asks why traffic doubled and bookings did not.

StepFinding (made up)Action
Country report2,350 sessions from one country with no patientsAnnotate the dates, exclude from client reporting by segment
Hostname900 sessions on a hostname that is not the practice's domainCreate a hostname filter, test for 36 hours, then activate
Staff visitsReception opens the site 80 times a week (about 320 over four weeks)Add the office IP as internal traffic, filter in Testing, then activate
EngagementRemaining sessions average 41 secondsTreat as real

Result: 6,400 less 2,350 less 900 less about 320 staff sessions leaves roughly 2,830 sessions, below September's 3,100, which now fits flat bookings. The owner reports "sessions: 2,830 after cleaning, 6,400 raw, cleaning rules listed", not a single clean-looking number. Showing the raw number and the rule is what keeps the report fair.

Test plan with pass and fail rules

  1. Baseline. Note 30 days of sessions, key events and leads. Pass: you have all three. Fail: key events are not set up, so mark your form submit or call click first.
  2. Find the suspicious block. Pass: you can describe it in one line, such as "2,350 sessions, one country, under 2 seconds". Fail: you only know "traffic looks high", so keep looking before filtering.
  3. Filter in Testing. Pass: after 24 to 36 hours the test shows the expected drop. Fail: nothing changes, so the rule does not match, such as a wrong IP or hostname.
  4. Activate and note the date. Pass: new data stays clean for a week. Fail: real people disappear, so set the filter to Inactive and recheck. Remember that active filters change data permanently going forward.
  5. Recheck against leads. Pass: sessions and key events move together again. Fail: still divergent, so look for bots that run full browsers and escalate to the CDN.

Who should skip what

  • Skip hostname and internal filters if your traffic is under a few hundred sessions a month. The noise is not worth the risk of a permanent mistake.
  • Skip the CDN step if your numbers are fine and the site is fast.
  • Do not use bot labels from any analytics tool, including a host's Bots card, as the only basis for billing a client or deciding a campaign. A label is only as good as the rules behind it. Ask the vendor how it labels, and confirm with leads.
  • If you must prove traffic to a client or advertiser, use the log and the lead counts together, not one dashboard.

How I'd approach this

I would not touch a filter until I have written down the suspicious block in one line and checked it against leads. Then I would set the internal traffic rule, test it, and activate. After that, I would add a hostname filter only if I see foreign hostnames. For reporting, I always give two numbers, raw and cleaned, with the rule, and I keep leads as the number that matters. If you host on a platform with its own bots card, such as Ploy, use it as a second opinion next to GA4, not as a replacement. If you want a custom log dashboard or a fix for what the logs reveal, Blur Studio builds that, see what a custom build includes.

First-week plan. Day 1: export 30 days of sessions, key events and leads. Day 2: run the six checks in the spot-the-pattern table and write one line per suspect block. Day 3: add the internal traffic rule in Testing. Day 4: add a hostname filter in Testing if needed. Day 5: activate the filters that passed and annotate the date. Day 7: compare sessions to leads and write the two-number line for your next report. For the monthly report format, see the client reporting template.

A closely related page covers the same ground from another angle: Website Heatmaps: How to See Why Visitors Leave.

For the next step, see Competitor Comparison Pages: Options for Getting Them Built.

Questions people ask

Does GA4 filter bot traffic automatically?

Yes for known bots and spiders, per Google. You cannot turn that off, and you cannot see how much was removed. Unknown bots and spam can still get through, so do not describe GA4 as filtering all bots.

Can I remove bot traffic from past data?

No. Data filters apply going forward, and an active filter is permanent. For past months, annotate the period and report raw and adjusted numbers side by side.

How do I know if a spike is bots or a real campaign?

Check whether leads moved too. Real traffic from a campaign usually brings key events, normal engagement time and countries you serve. Bots usually bring none of those. Always check country, hostname and engagement before filtering.

Should I block bots in Cloudflare?

Only the ones causing harm. Cloudflare's Bot Fight Mode is free but cannot be bypassed by your own WAF rules, so it may challenge a service you use. Review what is challenged after you switch it on and check Cloudflare's page for what your plan includes.

Why do my analytics numbers not match my server log?

They count different things. Most simple bots never run the analytics script, so they show in the log and not in GA4. Ad blockers and cookie refusals can hide real people from analytics but not from the log. See how to read the log for AI bots.

Does form spam count as bot traffic?

It is bot activity, but it shows up in your inbox and lead counts, not necessarily in traffic. If your form leads include junk, fix that first. See how to stop spam on your contact form.

What does Ploy do about bot traffic?

Ploy's analytics docs say the Bots card counts AI crawlers separately from agent visits. They do not describe filtering rules for the human totals or a data retention period, so ask Ploy before you quote cleaned numbers. See the Ploy pricing page for plans.

Is it safe to bill a client based on cleaned traffic numbers?

Do not use any single cleaned number for billing. Bot labels depend on the rules behind them. Use leads, calls and the raw and cleaned figures together, and write down the rules you applied.

More in this topic: Websites: simple stats and reports for owners and clients (9 guides) →

Sources

Ready for the next step?

A 30 minute call. If you do not need me, I will tell you.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…