Shopify Stopped Counting Bot Sessions. Your Conversion Rate Just Moved and Your Business Didn't

In late September 2026 Shopify changed how it counts sessions, filtering out more identified bot traffic than it used to. Plenty of stores opened their analytics the following week to find their conversion rate had jumped, sometimes substantially, with no change to the product, the traffic, the ads or the number of orders. If that is you, nothing improved. The denominator moved.

What actually changed

Conversion rate is orders divided by sessions. Shopify did not change how it counts orders. It changed what qualifies as a session, by getting better at excluding traffic it can identify as automated.

So the bottom of the fraction got smaller while the top stayed the same, and the result went up. A store doing 100 orders from 10,000 counted sessions was reporting 1%. If 2,000 of those sessions were bots that are now excluded, the same 100 orders against 8,000 sessions reports 1.25%.

That is a 25% improvement in a headline metric produced entirely by a definition change. The business did exactly the same amount of business.

The direction is worth being clear about, because it catches people out: removing junk traffic makes your conversion rate look better, not worse. If you went looking for a problem when the number moved, you were looking for the wrong thing.

Why this matters more than it sounds

A reporting change that makes a number look nicer seems harmless. It is not, for three reasons.

Every comparison across the change date is now invalid. Week-on-week, month-on-month, this-year-versus-last, any chart that spans late September is comparing two different metrics that share a name. If you have an automated report or a client dashboard doing that comparison, it is currently producing a fiction.

Decisions built on the ratio inherit the error. If your rule is to scale when conversion rate clears some threshold, that threshold just got easier to clear for reasons that have nothing to do with performance. Scaling into that is scaling on a measurement artifact.

It will be quietly blamed on something else. The uncomfortable version of this is the store that made a genuine improvement in the same window, sees conversion rate up, and credits the new landing page. Now a bad conclusion is baked into what the team believes works, and it will influence decisions for months.

The uncomfortable implication for your ad platforms

Here is the part that got the most attention, and it deserves care because it is easy to overstate.

If a meaningful share of sessions on your store were automated, those visits came from somewhere. Some of it is ordinary background noise on the open web: scrapers, uptime monitors, security scanners, SEO crawlers, AI training bots. That traffic arrives at every site whether or not you run ads, and it has nothing to do with your campaigns.

But some share of paid traffic across every ad platform is also not human. That is a known and long-standing feature of digital advertising, not a revelation. What changed is that the gap between what a platform reports as clicks and what your own analytics counts as sessions just got more visible, because one side of the comparison got stricter and the other did not.

The honest framing is this: you are not learning that your ads were fraudulent. You are learning that the two systems count differently, and you can now see the difference more clearly than you could a month ago. That is useful. It is not a scandal, and treating it as one will lead you to the wrong remedy.

The related and much more common problem, where platform-reported clicks far exceed the landing page views your site records, is covered in link clicks versus landing page views, and most of that gap has mundane explanations.

How to rebuild a baseline you can use

The goal is not to recover the old numbers. It is to get back to a state where a movement in a metric tells you something. Four steps.

1. Mark the date in your reporting.Whatever you use, annotate the change. A vertical line on a chart with a note saying "Shopify session definition changed" will save you and everyone who reads that chart a genuinely wasted afternoon in three months.
2. Stop comparing conversion rate across the boundary. Not adjusted, not normalised, just stop. There is no clean conversion factor, because the share of bot traffic was never uniform across channels or days.
3. Switch your reporting to metrics that did not change meaning. Orders, revenue, spend, and cost per order are all unaffected by how a session is defined. So is average order value. If you report on these, the change is a non-event for you.
4. Rebuild conversion rate forward. Two clean weeks after the change is a usable new baseline. Treat anything before it as historical trivia rather than a comparison point.

If you manage client accounts, send a short note before anyone spots it themselves. Explaining a confusing number in advance reads as competence; explaining it after they ask reads as an excuse. The framing that works is covered in how to explain tracking issues to clients.

The check worth doing while you are in here

There is a second-order question this change raises that is worth ten minutes, because the answer is occasionally alarming.

If bot sessions were reaching your store, were they also firing your tracking events? A crawler that executes JavaScript will trigger a PageView. Depending on how your events are configured, a determined one can trigger more than that.

Open your Events Manager and look at the ratio between your top-of-funnel events and your purchase events over the last few months. A PageView count that is wildly out of proportion to everything downstream, or a ViewContent volume that does not square with the traffic you believe you have, is worth understanding. Inflated upper-funnel events do not just make reports look odd. They pollute the audiences you build from them, and they give the optimisation algorithm a distorted picture of who engages with you.

The events that matter most are the ones you optimise toward, and those are usually protected by the fact that a bot is not going to complete a checkout. That is genuinely reassuring. Purchase and lead volume are the numbers to trust here, which is the same conclusion the rest of this article keeps arriving at.

Your conversion rate benchmarks are now wrong too

A smaller consequence, but one that will quietly misinform decisions for months.

Most stores carry around a sense of what good looks like. Two percent is fine, one percent needs work, four percent is excellent. Those numbers came from a world where sessions included bot traffic, and every published benchmark you have ever read was computed the same way.

Post-change, your own conversion rate is measured on a stricter denominator than the benchmark you are comparing it against. You will look better than your peer group without being better, and if the comparison drives a decision, such as concluding the site is fine and the problem must be traffic, that decision now rests on a mismatch.

The same applies internally. A conversion rate target agreed six months ago is easier to hit today, so hitting it means less. Worth revisiting any goal or bonus that hangs off the metric, and any A/B test still running across the change date, where the control period and the treatment period are no longer measured the same way. That last one is worth checking carefully: a test that spans late September may show a winner that is purely an artifact.

The general lesson, which is bigger than this one change

This is the third measurement change in 2026 that moved a lot of dashboards without moving any businesses, alongside the Shopify script tag deprecation and the ongoing consent and attribution shifts covered in consent mode and your tracking.

The pattern is always the same. A platform improves how it measures something, your numbers move, and because nobody wrote down what normal looked like beforehand, the movement is indistinguishable from a real change in the business. Teams then spend a week investigating a performance problem that does not exist, or worse, take credit for an improvement that did not happen.

The defence is unglamorous and it is the same every time. Know the relationship between what your platforms report and what your own systems record, and check it on a schedule rather than in a panic. If you know that Meta typically reports 15% more purchases than Shopify records, then when that figure becomes 40% you have learned something specific on the day it happens. If you have never measured it, every number is just a number, and you are permanently one platform update away from confusion.

The wider version of that argument, including what ranges are normal between which systems, is in how much tracking discrepancy is normal.

What to do this week

• Annotate the change date in every report and dashboard that shows conversion rate.
• Pull orders, revenue and spend for the four weeks either side. If those are flat, nothing happened to your business and you can stop worrying about the conversion rate movement entirely.
• Check your top-of-funnel event volumes for anything wildly out of proportion to your order volume.
• Write down, today, the current ratio between platform-reported purchases and actual orders. That single number is the thing that will tell you about the next change on the day it happens rather than a month later.

None of that takes long. The last one in particular takes about five minutes and is the single highest-value habit in ad measurement, which is why it turns up in nearly everything written here.