The chart of sessions from Google is falling, someone forwards an article on “the end of SEO”, and the board asks what the plan is. Before any decision is made, it is worth separating two things: what AI answers in search do to clicks, and what they do to orders. These are not the same phenomenon, nor the same scale. This guide shows how to measure your own case using Search Console queries and translate it into sales, instead of panicking over sessions.
Mechanism
What AI answers change in search results
The mechanism is simple: users get part of the answer above the results and click nowhere. Google changes the names and coverage of these modules (today these include AI Overviews), but the principle remains: the more a query can be “closed” with a paragraph of text, the less often it ends in a click. That is why the impact is distributed extremely unevenly between two types of query:
how-to query: “how to replace brake pads”
The user gets the instructions on the spot: impressions remain, clicks melt away.
product query: “brake pads [car model]”
Purchase intent requires a shop: offers, prices and availability still lead to a click.
Measurement
Measure your case: impressions versus clicks
The signature of the effect is visible in Search Console data per query: stable or rising impressions, falling clicks, and a drop concentrated in how-to queries. Split your queries into two groups and compare year on year (car parts shop, one quarter, illustrative data):
| Query group | Impressions YoY | Clicks YoY | Organic CTR | Interpretation |
|---|---|---|---|---|
| How-to (blog, “how to…”) | +9% | −34% | 4.1% → 2.5% | classic AI module signature |
| Products and categories | +3% | −4% | 3.4% → 3.2% | within normal fluctuation |
The same analysis per page (blog posts versus category and product pages) confirms the diagnosis from the other side. An important caveat: a uniform drop in impressions and clicks across all groups is a different case (rankings, seasonality, competition) with a different diagnosis, described in the guide on finding the causes of a decline.
Conversion
Sessions are not sales: measure the impact on orders
Panic comes from counting sessions; calm comes from counting orders. Sessions from how-to queries usually convert at a fraction of a per cent, product sessions many times better. The same quarter converted into estimated orders, using each group’s conversion rate from GA4 (blog pages versus product and category landing pages):
| Group | Lost sessions / qtr | Group conversion rate (GA4) | Lost orders | Lost revenue |
|---|---|---|---|---|
| How-to | −18,400 | 0.3% | −55 | −€3,300 |
| Product | −900 | 2.6% | −23 | −€1,375 |
| Total | −19,300 | – | −78 | approx. −2% of organic revenue in GA4 |
The takeaway from the example: 95% of lost sessions account for less than three quarters of lost revenue, and the entire effect is about 2% of organic revenue according to GA4, even though the sessions chart suggested a disaster. The reverse also happens: if product queries are melting away, the problem is serious and requires action. That is exactly why the calculation is done on your own figures, group by group.
The costliest overreaction
Cutting investment in content and SEO “because AI is taking traffic” without doing this calculation. If the decline sits in sessions converting at 0.3% while product queries hold steady, the shop is losing traffic of negligible value and sales remain untouched. Content strategy decisions are made on orders.
Defence
Defence: queries AI cannot settle in a paragraph
A shop holds its ground where a text answer cannot replace a visit to the shop:
- The product long tail: queries containing product, model and category names. Here the quality of product pages and the coverage of the range in the index decide the outcome, and the intent ends in the basket.
- The product feed as a second front: offer cards in search results (free and paid) draw on the feed rather than on page content. Refined titles and attributes, following the guide to Google feed fields, work independently of AI modules.
- Brand searched by name: branded queries are resilient because the user wants a specific shop. They are built outside the search engine; within search itself, it is worth monitoring who captures them (see the guide to search terms).
- Citable content instead of generalities: AI modules are more likely to point to sources with specifics: original measurements, tables, definitions, figures. A “10 ways to…” guide competes with the AI paragraph; unique data from your own business is what AI has to quote with a source attribution.
DataOrganizer
How DataOrganizer measures AI’s impact on your traffic
The whole calculation described here rests on three datasets that DataOrganizer keeps together: queries and pages from Search Console (impressions, clicks, CTR, positions), sessions, conversion rates and revenue per landing page from GA4, and paid orders from the shop as a check on the overall daily trend. Query segmentation, year-on-year comparison and conversion into estimated orders all fit into a single conversation:
Breakdown of the click decline and its cost:
| Query group | Clicks YoY | Group conversion rate (GA4) | Revenue impact |
|---|---|---|---|
| Informational (92% of lost clicks) | −34% | 0.3% | −€3,300 / qtr |
| Products and categories | −4% | 2.6% | −€1,375 / qtr |
The pattern points to AI modules: impressions for the informational group rose while CTR shrank, and 92% of the lost clicks converted at 0.3% according to GA4. The total impact is around 2% of organic revenue in GA4. Product queries are holding steady; recommendation: keep expanding category pages and the feed rather than cutting content, and assess informational guides by how they support sales.
Methodological note: the figures in the examples are illustrative, and the boundary between query groups can be fluid; what matters is applying the split consistently across periods. Google changes the names and reach of its AI modules, so the analysis should be repeated regularly rather than treated as a one-off verdict; the measurement mechanism itself (impressions versus clicks per intent group, converted into orders) remains valid regardless of the names.
Key takeaway
AI in search changes how clicks are distributed; SEO still makes sense for a shop. Informational queries lose low-value sessions, while product queries keep selling. Those who measure the impact on orders make decisions; those who measure it on sessions make panic.
FAQ
FAQ: AI in search and shop traffic
Is AI in Google taking traffic from online shops?
Unevenly: it hits informational queries hardest (the answer “closes” above the results) and product queries far less, as purchase intent requires visiting the shop. The scale varies by shop; it should be calculated on your own queries rather than taken from headlines.
How can I check whether this affects my shop?
Split your Search Console queries into informational and product groups, then compare impressions, clicks and CTR year on year. The AI module pattern: impressions stable or rising, clicks falling, with the decline concentrated in the informational group. An even decline across everything points to a different diagnosis (positions, seasonality).
Does a drop in clicks mean a drop in sales?
Not directly: the loss is calculated by group, multiplying lost sessions by each group’s conversion rate. Informational sessions convert at a fraction of a percent, so losing them usually costs little; only a loss in product queries really hurts.
How can a shop defend its traffic?
The product long tail and category pages, a refined feed (offer cards work independently of AI modules), a brand built outside the search engine, and content with specifics (own data, tables, definitions) that AI quotes with a source attribution rather than replaces.
How does DataOrganizer help?
Search Console queries and pages, GA4 traffic and conversion rates, and paid orders from the shop in a single model: intent segmentation, year-on-year comparisons and converting lost clicks into estimated orders using GA4 conversion rates take just a few chat questions, so the answer to “how much does this really cost” arrives in minutes.
Cheat sheet for the whole guide
AI answers in search: calculation instead of panic
1. Separate intent groups. Informational and product queries live in different worlds: the former lose clicks, the latter still lead to the basket.
2. Measure on orders. Lost sessions × group conversion rate = real cost. A sessions chart exaggerates the problem precisely where sessions are worth least.
3. Strengthen resilient fronts. Product and category pages, the feed, the brand and content with your own specifics. DataOrganizer keeps Search Console, GA4 and orders together, so the whole calculation takes one question and can be repeated every quarter.
DataOrganizer · organic clicks converted into estimated orders
Connect Search Console, GA4 and your shop: intent groups, year-on-year comparisons and conversion into orders in one place. Instant demo account, and you pay only for the sources you connect.