There’s a lot of noise about this and not much measurement. Search for it and you’ll find the same handful of percentages repeated across LinkedIn posts and agency blogs, none of them measured on a store like yours, several of them contradicting each other.
Your own data can answer the question properly, and it takes about an hour. Here’s how I’d do it.
First, separate the two things being blamed
They get talked about as one problem and they behave completely differently.
AI Overviews are a Google Search feature. When somebody clicks a link inside one, that click lands in your Search Console data as normal organic search. There’s no separate report for it. So the effect of AI Overviews shows up as a change in the relationship between impressions and clicks, never as a line item.
Assistants like ChatGPT, Gemini, Perplexity and Claude are different products entirely. If somebody arrives from one, it shows up in your analytics as referral traffic from that domain. That’s directly countable, and it’s the check I most often find nobody has run.
Any answer that treats these as one thing is going to be wrong about at least one of them.
The Search Console check that actually tells you something
Open Search Console, set the date range to the last 16 months, and compare impressions against clicks over time.
What you’re looking for is divergence. If impressions have held up or grown while clicks have fallen, something is answering the query before people reach you. If both have fallen together, that’s a ranking or a demand problem, and AI isn’t your culprit.
Then split it by query type. A store can get a clearer picture here than a lot of sites can, because the intent behind the queries is easier to separate. Filter to queries containing question words, “how”, “what”, “which”, “best”, “vs”. Those are the pre-purchase questions most likely to be answered on the results page. Compare that group’s click-through rate against your product and brand queries over the same period.
If the question-led group has lost click-through while your product queries are steady, you’ve found it, and you’ve found it in your own numbers rather than in somebody’s headline.
The GA4 check for assistants
In GA4, go to Traffic acquisition and filter session source for the assistant domains. chatgpt.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com will cover most of it.
Worth knowing before you look. The numbers are usually small, and they’re almost certainly understated, because some assistant traffic arrives without a referrer and lands in direct. So treat it as a floor, not a measurement.
The more useful move is to set the segment up now even if the figure is close to zero. You can’t backfill a trend you never recorded, and in twelve months you’ll want to know what it was today.
What to do about it depends on what you found
If the answer is that your question-led content is losing clicks, that’s genuinely worth acting on, and the action isn’t what most people sell you. There’s no file to add and no markup that makes an assistant prefer you. Google says plainly that no special files or markup are needed for its AI features.
What changes the odds is being the source worth quoting. For a store that usually means the pre-purchase content: the sizing guide, the compatibility answer, the honest comparison between two products you sell. Specific, verifiable, and better than the manufacturer’s copy that every other stockist is running.
If the answer is that nothing much has changed, that’s worth knowing too, and it should stop you buying an AI visibility package you don’t need yet.
The check worth doing yourself
Ask the assistants what they currently say about your products.
Write down the questions a customer would genuinely ask before buying from you, in their words, and run them through a couple of assistants in a fresh session. Record whether you’re named, whether the description is accurate, and which sources get cited instead of you.
Do it more than once, because answers vary between runs. What you’re building is a baseline you can compare against in three months, which beats any single screenshot.
If a competitor comes up and you don’t, look at what the cited sources actually contain before assuming there’s a technical fix. Often they’ve simply answered something specific that nobody else bothered to write down.
What I’d take from all this
The honest position is that this is measurable on your own store, and in my experience it usually hasn’t been measured. The percentages circulating are not about your catalogue, your categories or your customers.
Run both checks, set up the segment, take the assistant baseline. That’s an hour, it costs nothing, and it puts you in a position to tell whether any of this is worth spending money on.
If you want a hand with it, the AI search visibility page covers how I approach it, and Shopify SEO and WooCommerce SEO cover the platform side.
Questions people ask
Can I see AI Overview clicks separately in Search Console?
No. Google reports clicks from its AI features inside your normal Search performance data, with no breakdown isolating them. That’s a design decision on Google’s part and there’s no workaround, which is why the impressions against clicks comparison is the best available signal.
How much traffic do assistants actually send?
On the accounts I see, not much yet, though it varies by sector and I wouldn’t treat mine as representative of yours. That’s why the number worth having is your own rather than someone else’s. Set up the GA4 segment and you’ll know, and you’ll have a trend to compare against later.
Should I buy an AI visibility tool?
Check your own numbers first. If the pre-purchase questions are losing clicks and assistants don’t name you, there’s a case for doing the work. Whether that needs a tool or just better content is a separate question, and the tools mostly report on a problem rather than fix it.
Does schema markup help me appear in AI answers?
Google states that no additional markup or files are needed for its AI features. Schema helps search engines understand your products and is worth having for other reasons. Treat anyone selling it as an AI visibility solution with some caution.
What if my traffic is down but my rankings aren’t?
That’s the pattern worth investigating, and it’s what the impressions against clicks comparison is designed to surface. It can be AI answers, and it can also be a changed search results layout, more ads, or a competitor with a better title. Worth checking before concluding.
