Find and filter out bot traffic

Bot traffic gives itself away when a slice of your visitors behaves in a way no group of real buyers would. Kite looks for those slices by breaking your traffic down and comparing each piece with the period before. It reads visitors, sessions, bounce rate, and time on site, grouped by:

• traffic source and referring site
• country and city
• browser and browser version
• device type and, in Google Analytics, screen resolution
• landing page.

A real change in interest usually spreads across several of these. A bot campaign tends to pile up in one. Here are the patterns Kite treats as warning signs:

• A sudden jump in traffic from one city, especially a data center town such as Ashburn, Virginia, with sessions that last a second or two.
• A referring site you've never heard of that sends hundreds of visits, none of which go past the first page.
• Visits from a headless browser, which is a browser run by software with no screen, or from browser versions several years out of date.
• Direct traffic to a single page, such as your signup form, with no visits to the rest of your site.
• Traffic that arrives at the same minutes every hour or at a steady rate through the night.

In PostHog, Kite can also read the session recordings behind a suspicious slice. It doesn't watch the video. It reads each recording's length, starting page, and activity, so it can see when hundreds of sessions show no mouse movement and no scrolling. In Google Analytics, Kite can run a realtime report while a spike is still happening, which helps when the burst stops before the next day's numbers come in.

For each pattern it finds, Kite tells you how many visits it covers, the dates, and why it thinks those visits aren't human. When the evidence is mixed, it says so. A spike from one country might be a bot, or it might be a newsletter mention in that country, and Kite checks the referrers and landing pages before calling it either way.

Signups that never do anything come from one of two places: bots filling in your form, or real people who signed up and then got stuck. Kite tells the two apart, because the fix for one is useless for the other. Blocking bots won't help people who are confused by onboarding, and rewriting onboarding won't stop a script.

Kite checks what happened around each signup in your PostHog project. It looks at what the visitor did before signing up and what the account did afterward. These patterns point toward a bot:

• the signup arrived within seconds of the first page view
• the visitor never loaded the page the signup form sits on
• many signups arrived within a few minutes from the same browser and location
• the names are random strings, or the email addresses follow a machine pattern or come from throwaway email services
• the account never confirmed its email address or opened your product.

Real people who don't get started leave a different trail. They usually read a few pages first, often your pricing page, and many open your product once before they drift away. When Kite sees that pattern, it tells you the signups are real and points you to where they stop, such as the step after account creation. That turns "why do so many signups never do anything?" into a specific answer: for example, 140 of last month's 400 signups look like bots, and most of the real ones stall at connecting their first data source.

Once Kite knows which traffic and signups are fake, it rebuilds your funnel without them so you can see your real numbers. In PostHog, it builds the same funnel twice, once as your reports show it today and once with the suspected bots excluded. In Google Analytics, it runs your traffic and conversion reports with the suspicious sources, locations, or browsers left out.

The two versions side by side show exactly how much the bots were distorting your picture. A landing page that seemed to convert at 8 percent might convert at 3 percent once the fake form fills are gone. A campaign that seemed to bring in a flood of signups might have brought in a handful of real ones. Kite marks which conclusions change, such as which channel looks best, and which ones still hold.

Kite also checks whether the bots skewed your past decisions. If a bot wave started the same week you launched a new page, the launch may have looked like a success when it wasn't. In PostHog, Kite can add a note to your charts on the day the bot traffic started, so anyone reading the trend later knows why it jumped.

Filtering bots out of your reports for good works differently in each tool, and Kite does what each one allows.

In PostHog, Kite can mark each suspected bot by setting a property on that person's profile and collecting them into a cohort, which is a saved group of people. It then adds a rule to your project's filter for internal and test users, so every chart and funnel that uses that filter leaves those people out. Kite shows you the exact rule before it changes your project settings. Nothing is deleted. The tagged people stay in your project, so if Kite flagged a real customer by mistake, removing the tag puts them back in your numbers.

Google Analytics gives you fewer controls. It already removes known bots on its own, and it doesn't let you add your own rules for excluding them from collected data. Kite works around that in two ways. It runs your reports with the bot traffic excluded and gives you the corrected numbers. It also tells you which Google Analytics settings to change, such as adding a spam referral site to your list of unwanted referrals, and it explains what each change will and won't fix.

Bot traffic tends to come back, often from a new source or with a slightly different pattern. You can ask Kite to repeat the check on a schedule, such as every Monday. Each run compares the week with the one before, flags any new slice of traffic that looks automated, and reports how many signups looked fake. Results come to your Slack channel, by email, or to the Kite web app.

If your product sends signups to PostHog, Kite can also look at each new signup as it arrives. When a burst of bot signups starts, Kite tells your team right away instead of waiting for the weekly check. That gives you time to protect the form before a week of fake leads lands in your CRM.