Quick answer
A bot farm is a coordinated collection of automated accounts, scripts or devices run by one operator to imitate real users at scale, usually to fake clicks, visits, views, installs or followers. It differs from a botnet because the operator owns or rents the machines.
What is a bot farm?
A bot farm is a coordinated collection of automated accounts, scripts or devices, run by one operator, that imitates real users at scale. The purpose is usually to fake attention: clicks, visits, views, followers, installs or votes. Where one bot is a single program, a bot farm is the operation around many of them, including the servers, proxies, accounts and software that keep them running.
The terms around it are often confused, so it helps to fix them:
| Term | What it is | Who or what does the work | Typical aim |
|---|---|---|---|
| Bot | One automated program | Software | Any repetitive task, good or bad |
| Bot farm | A coordinated group of bots, accounts or devices under one operator | Software, often on rented servers or real devices | Fake engagement at scale |
| Botnet | Many infected computers or devices controlled remotely | Malware on other people's machines | Attacks, spam, fraud |
| Click farm | A bot farm or worker group focused on fake clicks | Scripts or paid people | Ad fraud, inflated clicks |
| Phone farm | Racks of real phones driven by software | Real devices | Fake installs and app engagement |
The main difference from a botnet is ownership of the machines. A botnet runs on devices its operator does not own; a bot farm usually runs on infrastructure the operator rents or owns. That matters for detection: a bot farm's traffic often comes from data-centre or proxy networks, while a botnet's comes from ordinary homes.
Owning a bot farm is not automatically illegal. What matters is what it is used for: testing your own site is legitimate, manufacturing fake reviews, fake ad clicks or fake political support is not.
What are bot farms used for?
- Fake social proof. Followers, likes, reviews and video views bought to make something look popular.
- Ad fraud. Fake clicks and impressions that drain a campaign budget or inflate a publisher's earnings.
- Metric inflation. Visits and app installs generated to hit a target or sell a product on the strength of numbers.
- Influence operations. In July 2024 the US Department of Justice announced the seizure of two domain names and the search of 968 social media accounts on X that it said formed an AI-enhanced bot farm run for a foreign government. The accounts used invented personas, were registered through private email servers, and were created with a software package the advisory named Meliorator.
- Scraping and probing. Bots that read content in bulk or test pages and forms for weaknesses.
The case above shows the scale and the technique: a few operators, hundreds of accounts, each looking like a person. The same logic applies to website traffic, only with sessions in place of posts.
How much traffic is automated?
Automated traffic is now a large share of the web. The 2025 Imperva Bad Bot Report, which covers 2024, found that automated traffic made up 51% of all web traffic, the first time in a decade it passed human activity, and that bad bots alone were 37% of all internet traffic, up from 32% the year before.
Two cautions. First, this is a vendor report based on the vendor's own network, so treat it as an order of magnitude, not a measurement of your site. Second, most of that traffic never reaches an analytics script, because a lot of it is crawlers and API calls that do not run JavaScript; what you see in your analytics is only the part that does.
Traffic signatures a bot farm leaves in analytics
Bots are cheap to run but hard to make look like a varied human audience. Look for signatures that appear together, not for any one:
| Signature | What you see | Why a farm causes it | How it can mislead |
|---|---|---|---|
| One source or campaign dominates | A single UTM source or referrer suddenly contributes a large share | Operators work from one script and one tag | A real campaign can also produce a spike |
| Spike with no cause | Sessions jump on a day with no launch or mention | Jobs are run in batches | Press or social mentions do the same |
| One page and one second | Almost all sessions are single page with a duration near zero | Bots load and leave | Slow pages produce short sessions too |
| No interaction | Page views but no scrolls, clicks or events | Scripts do not behave like readers | Content that answers at a glance |
| Identical devices | Same browser, screen size and language in thousands of sessions | One software profile cloned many times | A corporate network can be uniform |
| Flat clock | Traffic at the same level through all 24 hours | Machines do not sleep | Global audiences can be flat too |
| Network concentration | Many sessions from a few networks or hosting providers | Rented servers and proxies | Offices and mobile carriers also concentrate |
| Zero outcomes | Thousands of sessions and no conversions | Nobody is buying | New traffic sources start at zero |
No single row is proof. Three or more together, on the same segment, is a strong case.
Worked example: how much a bot farm distorts the numbers
Sample data: a site had 10,000 sessions last month and 120 conversions. A segment of 1,200 sessions came from one UTM source called promo-x. In that segment 98% of sessions viewed a single page, the median duration was 1 second, and there were no conversions. The other 8,800 sessions had 45% single-page sessions and carried all 120 conversions.
| Group | Sessions | Single-page share | Conversions | Conversion rate |
|---|---|---|---|---|
| Whole site | 10,000 | 51.4% | 120 | 1.20% |
Source promo-x | 1,200 | 98% | 0 | 0.00% |
| Everything else | 8,800 | 45% | 120 | 1.36% |
The raw conversion rate is 120 ÷ 10,000 = 1.20%. Without the suspect segment it is 120 ÷ 8,800 = 1.36%, so the farm understated the conversion rate by 0.16 points: the reported figure is 12% below the real one (1.20 ÷ 1.36 − 1). If you pay for this source, 1,200 sessions bought nothing.
Count the signatures in the table: one source dominating (12% of sessions), single-page and near-zero duration, and zero outcomes. Three are present in the same segment, which is enough to treat it as suspect, and the other five (spike, interaction, device uniformity, clock, network) are worth checking before you block anything. A single spike on its own would not be enough.
Why detection is hard: farms that look like people
Detection works well on simple bots. Imperva's report names the tactics farms use to avoid it: residential proxies (so the traffic comes from ordinary home addresses), faked browser identities, headless browsers, AI-assisted scripting and anti-detection tools. A phone farm goes further by using real devices, so device checks pass.
That has a consequence for how you read your data. Detection of automated visitors can reduce the problem but never prove that the remainder is human. What cannot be faked cheaply is behaviour that costs money or effort: reaching a pricing page and then completing a purchase, signing up and returning on three separate days, a chat that makes sense. Judge segments by outcomes, not by labels alone.
What to do when you suspect a bot farm
- Isolate the segment. Find the source, campaign, landing page or network that carries the signatures, and look at it as a group.
- Check the outcome, not just the shape. If the segment has conversions or real conversations, it is not a farm, even if it looks unusual.
- Keep it out of the decisions. Report numbers with and without the segment so your conversion rate and bounce rate are not distorted.
- Stop paying for it. If the source is paid, pause it, and read the platform's policy on invalid traffic before disputing the charge.
- Block at the edge if it is hitting something sensitive. Rate limits, challenges and firewall rules belong at your CDN or web application firewall; analytics tools measure, they do not block.
- Write it down. Note the dates and signatures, so you recognise the next one faster.
Spotting a bot farm in TapCub
TapCub measures traffic; it does not block it. What it does is keep recognised bots out of your human numbers and let you look at suspicious segments.
- Check what is already excluded. New projects use Detect and report separately: recognised bots appear only under Web analytics › Bots and are not counted in human visitors, sessions or pageviews. All bots lists each one with its operator, purpose, requests, pages visited, most common status code, trust level and the detection layers it matched. See how bot detection works and are bots counted in visitors.
- Find the segment. Open Web analytics › Sources, scan the UTM source and referrer rows for the one that has jumped, and click it to add it as a filter. Only a hostname filter carries from one page to the next, so re-apply the source filter on Pages, Audience and Sessions (or keep the address that already has it) to check single-page share, duration, devices and entry pages for that segment alone.
- Check outcomes. Under Web analytics › Goals, see whether the filtered segment has any conversions, and compare with the whole site, as in the worked example.
- Set an alert. In All-platform analytics › Operations › Monitors, create a check on a built-in metric such as sessions, bounce rate or pageviews at an hourly or daily interval. A trial run shows how many times it would have fired over the last 30 days. Monitors only notify.
- Know the limits. With the page script alone TapCub sees only bots that run JavaScript, so crawlers that do not need server collection or a log import. Trust is verified, suspected or unknown, and unknown does not by itself mean a bot is spoofed. If a row is wrong, report a false positive.
For blocking, use your CDN or firewall. TapCub reports what happened; it does not stop it.
Sources
- Imperva, 2025 Bad Bot Report: How AI is supercharging the bot threat, vendor report covering 2024, checked 2026-11-20.
- US Department of Justice, Justice Department leads efforts among federal, international and private sector partners to disrupt Russian government-operated social media bot farm, July 9, 2024, checked 2026-11-20.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Experience Specialist
Writes about product experience, practical workflows, and growth, turning complex processes into clear steps for everyday decisions.