What drives search traffic, and which part you control

When search traffic rises or falls, three things drive it. Something changed on your side, either on your site or in how often your customers search. Something changed on a competitor’s side, on their site or in how they’re seen elsewhere. Or search engines and AI assistants changed how they choose and count results. Only the first is yours to change. A competitor’s pages are public, so you can see what they did, even when it’s hard to tell why it worked. With the platforms you see only the results changing, because the reasons behind them stay private. Most search traffic reports show all three as the same moving line, even though each of these drivers calls for a different response. Mistaking one for another usually means changing your site when it wasn’t the problem, or leaving it alone when it was.

Start with the part you control

Anything you change on your site can move traffic in either direction. A new page about a service you’d never written about can bring in searches you didn’t appear for before. A redesign that renames or removes pages, or a setting that tells search engines to skip a page, can take searches away. Either way, the pattern has a place and a date. The movement gathers on the pages that changed, and it starts within days or weeks of the change.

This is the one driver where you hold the “before.” You know what changed, on which page, and when. So the useful question is whether you knew where those searches stood before you made the change. With that starting point, the result becomes clear within a few weeks. Without it, a drop after a redesign could be the redesign, a competitor or an update, and there’s no going back to find out.

Your market belongs on this side too. You can’t change how often customers search, but you can know its shape in advance. A landscaping company sees searches thin out every November, and a pool company sees them climb every April. Google’s troubleshooting guide lists shifting demand, from seasonality or a new trend, as one reason traffic falls. The tell is in two numbers Google Search Console reports separately. Impressions count how often you appeared, and position shows where you sat. When demand falls, impressions and clicks drop together while position holds. You stayed in the same place, and fewer people came looking. A dip at the same time last year confirms it, and the response is to plan around the season.

A competitor’s gain comes out of your results

Page one has a fixed amount of space, and an AI answer cites only a handful of sources, if any at all. When a competitor earns a place, someone else loses one. You can do nothing wrong and still slip, because the change that moved you happened on someone else’s site.

The pattern is a slide concentrated in one topic, meaning a group of searches about the same thing, while nothing on your own pages has changed. A landscaping company might see every search about patios slip together, from what a patio costs to who builds them nearby, while its lawn care searches stay where they were. Competitors usually go after one service at a time, so their gains land on a topic rather than spreading across your site. Impressions can stay level for a while, then fall as you drift toward page two. Because you hold the “before” for your own changes, you can rule them out first, and that’s how you know to look elsewhere.

The tell is who sits above you now, and Search Console can’t show it. It reports only the searches where your own site appeared, so you see your position fall without seeing what replaced you. Even with their page in front of you, working out why it won takes some reverse engineering. Sometimes the answer is on the page, such as a question it answers that yours doesn’t. Sometimes it lies elsewhere, in their reviews or in who links to them, and some of it can’t be seen from outside at all.

Whatever you find, the response runs back through the part you control. You decide whether those searches are worth contesting, and if they are, the work happens on your own pages.

When the platforms change how they choose and count

Changes made by search providers are the easiest driver to blame and the hardest to confirm. Google says its results shift constantly as the web itself changes, and it announces only its notable updates, so almost any drop can be pinned on Google and almost none can be proven. In early August 2026, the tools that track rankings across thousands of searches showed heavy movement for days while Google’s status dashboard listed nothing. Something changed, and from the outside there was no way to say what. AI assistants announce even less. The steps they follow to search and choose sources hold steady, while which index they use and how much of each page they read change often, and mostly without notice.

When Google does announce an update, the description runs to a line, usually that it’s a regular update meant to surface relevant, satisfying content. The details of what changed and why stay private. Google’s head of Search set out the principle in a sworn statement in January 2026, saying that fighting spam depends on obscurity, because signals lose their value once outsiders know them. Detail about what an update rewards would reach people trying to game it just as surely as businesses trying to understand it. From the outside, you can see that results changed but not why.

A change in how results are chosen shows up as movement across many unrelated searches at once, starting close to a confirmed update. The response is usually restraint. For small drops in position, Google advises against radical changes to a page that’s already doing well, and it’s worth waiting for an update to finish before comparing anything.

The results page itself changes too. An AI Overview can appear above a search you rank for, leaving your position the same while clicks fall. That is the pattern behind traffic falling while leads hold.

AI answers also change from one check to the next. For one business we monitor, we asked AI assistants the same six questions in July and again in September. Three came back with different answers, and on two of those, the business’s site was no longer cited. That variation is normal. Each time a question is asked, the assistant searches, gets a shortlist, and picks its sources again, so the same question can cite you one week and a competitor the next. In search you hold a position, and in AI answers you hold a probability, as we explain in how to find out whether AI answers name your business. A single lost citation says little on its own, and a pattern across repeated checks is what shows the probability moving.

Then there is counting, which moves the numbers without moving anything else. Two recent changes affect nearly every site:

• September 2025. Google quietly stopped supporting a setting that rank-tracking tools used to load 100 results at a time. Reported impressions fell and average positions looked better, while the results people saw stayed the same. Google didn’t list it as a reporting problem, and industry coverage traced the drop to that change.
• May 2025 to April 2026. A logging error distorted impression figures for that whole period, and fixing it lowered them. Clicks were unaffected, and the old data won’t be corrected. Google records this one on its Data anomalies in Search Console page, which lists known reporting problems.

As a result, this September’s impressions set against last September’s compare an overstated number with an accurate one, while clicks compare cleanly. The tell for a counting change is impressions moving while clicks stay put, and the response is to do nothing.

All three move at once, all the time

The three drivers don’t take turns. In any given month, your own changes, a competitor’s work and the platforms’ decisions are all moving your results together, and none of them waits for you to look. That’s why separating them depends on watching continuously rather than investigating after a drop. You need a before and after on the same searches, with impressions, position and clicks kept apart, plus a view of who else is on page one. Assembled after the fact, that picture is usually missing the “before.”

DepthFinder compares whatever timeframe you’re viewing with the one before it, on the same searches, and calls out the searches that moved most. Each comes with a verdict on whether the movement is good or bad. Longer trend views show whether a change is a dip or a direction. You can see which competitors appear on page one for searches where you don’t. AI answers are measured on the same topics, and when you tell DepthFinder a page has changed, those answers are checked again, so the after is measured against a known before. That lets you tell which movement came from the part you control, so the changes you make land where they can help and can be measured once they do. See how DepthFinder tracks what moved.

Guessing wrong costs more than waiting

Each wrong guess has a price. Blame an update and you rewrite a page that was doing fine. Wait for Google to reverse course and a competitor keeps the place their better page earned. Treat a reporting fix as a collapse and a week goes into a problem that doesn’t exist. Knowing which driver you’re looking at before you touch anything matters more than keeping up with every change Google makes.