Search Traffic Down, Revenue Flat: You Lost the Visits That Never Bought

Search traffic is falling, and revenue has barely moved. Should you go and win the lost traffic back? As long as you are looking at the click count graph, that question has no answer.

TL;DR#

• When an AI summary appeared in the search results, 8% of visits went on to click a regular result link. When no summary appeared, the figure was 15%[1]
• If the visits you are losing came to look something up, and the visits that came to buy are still there, revenue holds even as traffic falls
• In a month traffic falls, the number to open is not the click count but RPS (revenue per session: revenue ÷ sessions)
• Three exits: RPS up with revenue held is healthy, RPS flat with revenue down is real damage, RPS up with revenue down is shrinking scale
• Google Search Console goes as far as impressions and clicks. It does not show how much revenue a given keyword produced

1. Traffic Can Fall by a Third and Revenue Stay Flat#

You open Google Search Console and search clicks are clearly down from last month. The revenue graph, meanwhile, is almost horizontal. Same month, same site, and only one of the two has dropped.

How search results look has changed over the past year. Pew Research Center examined the browsing records of 900 US adults and found that when an AI summary was shown in the search results, only 8% of those visits went on to a regular search result link. When no summary was shown, the figure was 15%[1]. The data is from March 2025 and covers general search rather than any one category. These are not figures for EC in Japan, but the direction is readable: on result pages where a summary appears, fewer visits continue through to a link.

Google, for its part, says it sends billions of clicks a week to sites through its AI features[2]. This article does not settle which of the two is closer to the truth, because a market-wide total and what is happening on your own site are separate things (Billions of AI search clicks a week: the market total isn't your site).

What is certain is that as more questions get answered at the top of the results page, the visits that end there satisfied are the first to disappear. Someone who arrived on "what is X" or "how to X" reads the summary and bounces. Someone searching a product name or a brand name, or checking price, shipping or stock, reads the summary and opens the site anyway. The visits that leave and the visits that stay are different in kind, which is how you end up with traffic down by a third and revenue down by 2%.

Worth noting here is that none of this proves AI search is the cause. Seasonality produces the same shape. So does a competitor entering, or a change in the products you carry. Rather than settling on a cause first, the order of work is to confirm what is happening on your own site.

2. Switch the Basis for the Call From Clicks to RPS#

The graph to open in a month traffic falls is not the click count. It is RPS (revenue per session).

RPS = revenue ÷ sessions

If revenue holds while sessions fall, RPS necessarily rises as a matter of arithmetic. Read the other way around: if RPS has not moved while traffic alone is down, then what you lost includes visits that were producing revenue.

There are three exits.

• Healthy — RPS is up and revenue is holding. What you lost were visits that were never going to buy, and no real damage has occurred
• Real damage — revenue is down and RPS is flat or falling. Visits that came to buy are being lost too
• Shrinking scale — RPS is up but revenue is down. Efficiency per visit improved, but the denominator is falling faster than the efficiency gained

The same "traffic fell" points the next move in opposite directions depending on the exit. If it is healthy, there is no hurry to win the traffic back; thickening the landing experience for the visits that remain does more. If it is real damage, the work is restoring the composition of the traffic. If it is shrinking scale, the work is widening the entry points whose efficiency improved. Look only at the click count and all three read as the same "it fell."

The same thing happens when traffic is rising. If traffic is growing while revenue isn't following, RPS is falling (Traffic up but revenue flat). Rising or falling, the call is made on the same single metric.

3. Find Out Which Search Keywords Did the Falling#

Once RPS has given you the direction, the next question is what exactly fell. Take the query list in Google Search Console and read it split into look-it-up keywords and buying keywords.

The first group takes forms like "what is X," "how to X," "how to calculate X." The second group is product names, brand names, and forms carrying words like "price," "shipping" or "in stock." If only the first group has fallen sharply while the second is holding, then what section 1 described — that the visits leaving and the visits staying are different in kind — is happening on your own site as well.

We see this shape on our own site. Over the last 30 days, the Japanese keyword with the most impressions was one asking for the definition of a term, and although the average position sits in the eighth spot, on the first page of results, the click-through rate came in under 0.1%. On the English side, definitional keywords pile up impressions while the click-through rate stays at 0%. It isn't that the ranking is poor. The ranking is good and the clicks still aren't happening.

That said, none of this establishes that "the AI summary took them." The same figures appear when the top of the results page is occupied by some other element, and when the search was only ever a request for a definition that rarely produces a click (Same rank, fewer clicks). What you can conclude reaches only this far: a keyword with plenty of impressions is not necessarily an entry point for revenue.

And this is as far as Google Search Console goes. What it produces per query is four things — impressions, clicks, average position and click-through rate[3] — and how much revenue the visits from that keyword produced is not shown. A performance report for generative AI search has been added, but what it shows there is impressions as well[4]. A study that tracked 100 blogs over four years reported a median search traffic decline of 85%[5], so the change on the traffic side has been running for a while (100 blogs, an 85% traffic drop).

And this call cannot be made on site-wide RPS. The site-wide figure mixes branded search and returning customers together, so which entry point fell and which entry point gained efficiency does not come out unless you look at them separately. Only when sessions and revenue sit side by side for each traffic source does it become clear which of the three exits you are at.

RevenueScope solution

RevenueScope displays sessions, revenue and RPS on the same screen for the same period. A comparison against the prior period is attached to each metric, so in a month traffic falls you can read which way RPS moved without recalculating anything.

The channel breakdown shows sessions, revenue and RPS for each traffic source — search, AI, ads, social and the rest. Site-wide RPS mixes branded search and returning customers together, so which entry point fell and which gained efficiency only becomes visible once they are separated. The problem of branded search obscuring true strength is covered in a separate article (Traffic up but no new customers).

On the search keyword side, estimated revenue is attached to each keyword by multiplying Google Search Console clicks by RPS for search traffic. Among the keywords whose clicks fell, which ones were entry points that produced revenue can be separated out, within the limits of an estimate.

What the Healthy Shape Looks Like#

Take Fictional Store A, with three months ago set against this month.

Note: Fictional Store A's figures are a teaching example, rounded for explanation. The demo screen runs on the sample store's sample data (refreshed daily), so the values shown there will not match this table.

Sessions fall from 10,000 to 7,000, while revenue stays roughly where it was. RPS rises from ¥100 to ¥140, and CVR moves from 2% to 2.8%. Of the three exits this is the healthy one, a state where only visits that were never going to buy have gone away. Read that far and the decision to put budget into winning traffic back can wait.

FAQ#

Frequently asked questions#

Q. If RPS is up, can we leave falling traffic alone?

A. For that month, yes. Whether the same shape continues into the next month is a separate matter. If revenue starts falling while RPS stays up, you have moved into shrinking scale, and from there the call switches to widening the entry points. Don't treat it as a one-time check — record the same two values side by side every month.

Q. How much does RPS have to move before it counts as "up"?

A. When the denominator is small, a few percent of movement happens easily. In a period with fewer than a few hundred sessions, a single order added or lost swings RPS a long way, so don't decide on which ratio is larger. Until enough volume accumulates to decide on, watching the session count and the order count themselves is the surer route.

Q. Our search traffic is falling too. Is AI search the cause?

A. That can't be answered without looking at your own data. Seasonality, a competitor entering, a change in products, a site rebuild — all produce the same shape. Before settling on a cause, the order of work is to confirm which search keywords fell and whether those keywords were producing revenue.

Q. Will the generative AI search report in Google Search Console tell us?

A. It will tell you impressions on AI search surfaces. What appears there, though, is impressions[4], not how much revenue came from the clicks. It is material for understanding the change on the traffic side, but deciding whether real damage has occurred needs figures from the revenue side.

Summary#

In a month search traffic falls, acting in a hurry on the click count graph alone means spending budget to win back visits you were better off losing. On result pages where an AI summary is shown, fewer visits continue through to a link[1], and the ones that go first are the look-it-up visits. If the visits that came to buy are still arriving, traffic can fall by a third and revenue barely move.

To tell the two apart, look at which way RPS — revenue divided by sessions — has changed. Up with revenue holding is healthy. Flat with revenue falling is real damage. Up with revenue falling is shrinking scale. The next move differs across the three. And because this change is still under way, one check is not the end of it.

The basis to judge by in a month traffic falls is not the click count. It is where RPS is heading.

References#

• [1] Pew Research Center "Google users are less likely to click on links when an AI summary appears in the results" (2025)
• [2] Search Engine Roundtable "Google: We Send Billions Of Clicks To Websites Weekly Through AI Search" (2026)
• [3] Google Search Console Help "Performance report (Search results): Overview and basic setup" (2026)
• [4] Google Search Console Help "Generative AI performance report (Search)" (2026)
• [5] Daniel Stanica (Monetize Better) "The Great Blogging Collapse" (2026)