How to tell whether your traffic drop is a real problem

You opened the numbers because something felt off, and there it is: visits down by a third against last year, falling steadily since spring. The obvious conclusion is that the business is in trouble. Then you check the other number and find that enquiries are flat, jobs are flat, and revenue is flat or slightly up. Two lines that moved together for a decade have come apart, and only one of them looks like bad news.

A visit stopped standing in for a customer

For twenty years a visit to your website was a reasonable proxy for interest. Somebody wanted to know something, so they came and read it, and counting visits counted interest. That link is what has broken. People are still asking the same questions; a growing share of them now get the answer without ever arriving to collect it.

It is a strange kind of bad news, because nothing you built is broken. Your ruler is.

What the research actually shows

The most careful study available comes from the Pew Research Center, which followed what 900 people in the United States actually did across 68,879 Google searches rather than asking them about it afterwards. When an AI summary appeared, people clicked a normal search result on 8% of visits. When no summary appeared, they clicked on 15% of visits, close to twice as often. Links inside the summary itself were clicked on 1% of visits.

Separately, Ahrefs compared 300,000 searches and found 34.5% fewer clicks for the top result when an AI summary was present. Their own warning is worth repeating, because it is more useful than the number: nobody can yet separate these clicks out cleanly, including them. Treat all of this as a rough shape rather than a measurement of your business.

A word about the numbers you will be shown

You will see a 58% figure attributed to that Ahrefs study in a great many articles. It is not the number on their page. This whole subject is full of statistics that have drifted away from wherever they started, growing rounder and more alarming with each retelling, and the drift always runs in the direction that makes somebody's product look more necessary.

Before acting on a statistic about this, click through to the study and check the number is really there. It takes a minute and about half the time it is not.

Work out which of your numbers is broken

There are three patterns and they call for three different responses, so it is worth being honest about which one you are in before spending anything.

If visits are down and jobs are flat, your measuring changed rather than your business. Do not rebuild the website and do not spend money solving it; write down what happened so the same graph does not frighten you again next quarter. If visits and jobs are falling together, something real is going on, and it is far more likely to be a competitor, a season or a price than anything to do with AI. And if visits are flat while jobs fall, people are arriving and leaving without contacting you, which is a website problem and much the most fixable of the three.

Start asking people again

The cheapest instrument you have is a single question on your enquiry form, or asked on the phone: how did you hear about us? It fell out of fashion because software was supposed to answer it for us, and software has largely stopped answering it. A tally on a sheet of paper for one month will tell you more about where your work comes from than any dashboard currently can, and it produces answers no report would ever show you, like somebody's sister recommending you or somebody asking an app.

The part that is easy to miss

If fewer people read your site before they choose, then what is written about you matters more rather than less. It is simply being read by something that shortens it first. Pew found the median AI summary is 67 words long, and that 88% of them point at three or more sources.

Sixty-seven words is not much room for a maybe. So the work comes back round to something dull and useful: saying plainly what you do, where you work, when you are open and what it costs.

Sources

• Pew Research Center — 900 US adults, 68,879 searches, March 2025; clicks of 8% with a summary against 15% without.
• Ahrefs — 300,000 keywords; 34.5% lower average click-through for top-ranking pages, with their measurement limits stated.

Written by the team at hailera — we help small businesses answer customer questions without being there.