Conversion rate is a fraction. The top of it, orders, comes from your store database and is about as reliable as data gets. The bottom of it, sessions, comes from a JavaScript tag that runs in whatever loads your pages. That includes shoppers, but it also includes scrapers, price monitors, uptime checkers, AI crawlers and automated browsers. When automated sessions grow, the fraction shrinks, and nothing about your store changed.
This is no longer a fringe problem. Imperva’s 2026 Bad Bot Report found that bots generated 53 percent of web traffic in 2025, up from 51 percent the year before. Most of that never reaches your analytics, but some does. In a study of 787 stores tracked between late July and October 2026, Shero Commerce and Littledata found that 133 of 687 Littledata stores (19 percent) and 37 of 100 Shero stores (37 percent) recorded 5 percent or more of their GA4 sessions from bots. On the affected stores, the median real conversion rate was 10 to 14 percent higher than GA4 reported.
Ten percent is the difference between a checkout redesign that looks like it failed and one that worked. This guide covers what GA4 already removes, what slips through, how to size the problem on a WooCommerce or Magento store, and how to report a conversion rate built on human sessions.
What GA4 removes, and what it does not
GA4 has a built-in filter, and it is always on. Google’s documentation says known bot traffic is identified with its own research and the IAB International Spiders and Bots List, and that you cannot disable known bot traffic exclusion or see how much was excluded. The filter works on bots that announce themselves. A scraper that pretends to be Chrome is not on any list.
How much gets through depends on the bot. Clickport, an analytics vendor, sent 1,000 automated sessions to a test site running GA4 in March 2026, using five setups from plain scripts with bot user agents to stealth headless browsers on residential proxies. In its write-up, GA4 counted all 1,000 sessions. That is one property and a vendor running its own test, so treat it as an illustration rather than a miss rate. The broader point holds: any automation that executes your GA4 tag and is not on the known list becomes a session.
| Type of automated traffic | Runs your GA4 tag? | Removed by GA4? | Where you can see it |
|---|---|---|---|
| Search engine crawlers (Googlebot, Bingbot) | Sometimes | Yes, on the known list | Server logs |
| Simple scrapers and scripts (curl, python-requests) | No | Not needed | Server logs only |
| Headless browsers and price monitors | Yes | Usually not | GA4 and server logs |
| Referral spam and fake Measurement Protocol hits | Sent straight to GA4 | Usually not | GA4 only |
| AI crawlers and shopping agents | Varies by agent | Only if listed | Server logs, sometimes GA4 |
How much bot traffic moves your conversion rate
Bots add sessions and almost never add orders, so the correction is simple. Real conversion rate equals reported conversion rate divided by the share of sessions that are human. Shero’s example: a store with 100,000 sessions, 20 percent of them bots, and 1,600 orders shows 1.6 percent in GA4 while shoppers convert at 2.0 percent. The table runs the same arithmetic for other bot shares.
| Bot share of GA4 sessions | Reported conversion rate | Real conversion rate | Understated by |
|---|---|---|---|
| 2% | 2.00% | 2.04% | 2% |
| 5% | 2.00% | 2.11% | 5% |
| 10% | 2.00% | 2.22% | 10% |
| 20% | 2.00% | 2.50% | 20% |
| 35% | 2.00% | 3.08% | 35% |
The store-wide number is the least of it. Bot traffic is not spread evenly, so it distorts the segments people make decisions on. The same study found that 95 to 96 percent of bot sessions were recorded as Direct traffic, which makes Direct look like your worst channel and inflates the “everything else” bucket in attribution. Landing page reports, country reports and A/B tests get the same treatment: a variant that happens to receive more scraper hits looks like it converts worse.
Bot share also changes over time. A new scraper finding your catalogue can add thousands of sessions in a week, and the conversion rate drop that follows looks exactly like a site problem. If you have chased a sudden fall in conversion rate with no matching fall in orders, this is a likely cause, and it sits next to the other gaps covered in our guide to why GA4 revenue never matches your WooCommerce or Magento orders.
Step 1: Anchor on orders, not sessions
Start with a daily chart of GA4 sessions next to orders from your store database. Orders are the one number automation cannot inflate in normal conditions. Nostra’s investigation guide makes the same point and treats a step change in sessions that never shows up in revenue as the first red flag. Look for three patterns:
- A step up in sessions with flat orders. Traffic jumps on a specific date and stays there. Real demand rarely arrives that way.
- Weekday shape disappears. Human traffic follows the week. Automated traffic often runs at the same level every day, so a flattening weekly pattern is a clue.
- Engagement rate falls as sessions rise. Sessions up, engaged sessions flat means the new sessions are not doing anything.
Step 2: Build a likely-bot segment in GA4
No single signal identifies a bot, but several together get close. Build an exploration or a BigQuery query that splits sessions by the signals below, then look for groups with plenty of sessions, zero engagement and zero orders.
| Signal | What it looks like | Watch out for |
|---|---|---|
| Engagement | Not engaged: under 10 seconds, one page, no key event | Real bounces from slow pages look the same |
| Geography | Countries you do not ship to, data centre cities such as Ashburn, Virginia | Shoppers on VPNs |
| Landing page | Deep pagination, internal search URLs, one product page hit repeatedly | Shared product links |
| Browser and OS | Linux desktop, outdated Chrome builds, one exact build overrepresented | Developer and staff traffic |
| Funnel ratios | Add-to-cart events with no checkouts, repeated from the same source | Shoppers saving carts |
| Source | Direct, or referrers you have never heard of | Untagged email and app traffic |
If you export GA4 to BigQuery, this query gives sessions, engaged sessions and purchases by country and browser for a month. Sort by sessions and look at the rows where purchases are zero and engaged sessions are a small fraction.
-- GA4 BigQuery export: sessions, engaged sessions and purchases by country and browser
WITH s AS (
SELECT
CONCAT(user_pseudo_id, CAST((SELECT value.int_value FROM UNNEST(event_params)
WHERE key = 'ga_session_id') AS STRING)) AS session_key,
geo.country AS country,
device.web_info.browser AS browser,
MAX((SELECT COALESCE(value.string_value, CAST(value.int_value AS STRING))
FROM UNNEST(event_params) WHERE key = 'session_engaged')) AS engaged,
COUNTIF(event_name = 'purchase') AS purchases
FROM `your-project.analytics_123456789.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20261001' AND '20261031'
GROUP BY session_key, country, browser
)
SELECT country, browser,
COUNT(*) AS sessions,
COUNTIF(engaged = '1') AS engaged_sessions,
SUM(purchases) AS purchases
FROM s
GROUP BY country, browser
ORDER BY sessions DESC
LIMIT 50;
Step 3: Check the server logs GA4 never sees
GA4 only sees automation that runs JavaScript. Your web server sees every request. On a WooCommerce site that usually means the Nginx or Apache access log, often exposed by the host; on Magento it is the same plus whatever your CDN keeps. Group requests by user agent and IP range for a day and you will usually find a handful of sources making a large share of requests. Two cautions. User agents are easy to fake, so verify any request claiming to be Googlebot with a reverse DNS lookup, as Nostra recommends, rather than trusting the string. And logs measure load, not analytics distortion: a scraper that never runs your tag costs server capacity but does not touch conversion rate.
This step also catches AI crawlers. The traffic they send back when a shopper clicks a cited link is valuable, and we covered how it reaches GA4 in our post on ChatGPT referrals filed under Direct. The crawling itself is a separate stream and belongs in the logs, not in your conversion rate.
Step 4: Report conversion rate on human sessions
Shopify now labels sessions itself: its reports have a human or bot session dimension for comparing the two conversion rates. WooCommerce and Magento stores have no equivalent, so you build it. Two practical options:
- Engaged-session conversion rate. Divide orders by engaged sessions. GA4 counts a session as engaged if it lasts more than 10 seconds, has a key event or has two or more page views. This is easy to set up and removes most single-hit automation. It also removes real one-page bounces, so the number will be higher than any industry benchmark; compare it with itself over time.
- Exclusion segment. Take the groups you flagged in Step 2 and exclude them with a GA4 segment or a filter in your reporting layer. Start the filter in testing mode, because GA4 data filters are not retroactive and a wrong filter deletes real data going forward.
Keep the raw session count too. It is the right denominator for server cost and for anything billed per visit. The point is to stop using it for decisions about the shopping experience.
What filtering does not fix
Cleaning reports changes what you see, not what happens. The bots still load pages, still consume capacity during sales events and still collect your prices. Nostra also points out that automated load can slow pages for real shoppers, which lowers the real conversion rate as well as the reported one. If the volume is large, blocking belongs at the CDN or firewall. The analytics fix and the blocking fix are separate jobs, and the first one is cheap enough to do this week.
A short checklist
- Chart daily sessions against daily orders for the last 12 months and mark every step change.
- Split October and November sessions by country, browser and landing page; flag groups with zero orders and low engagement.
- Pull one day of server logs and list the top 20 user agents and IP ranges by request count.
- Add an engaged-session conversion rate to your weekly report next to the standard one.
- Before judging any A/B test, check that bot share was similar in both variants.
Sources
- Shero Commerce and Littledata, How to Stop Bot Traffic on Your Online Store (October 2026)
- Google Analytics Help, Known bot-traffic exclusion
- Clickport, GA4 Bot Traffic Test: 1,000 Sessions in March 2026
- Nostra, Bot Traffic in Google Analytics: How to Find and Filter It (August 2026)
- Imperva, Bad Bot Report 2026: Bots in the Agentic Age
- Shopify, Bot traffic detection


