Why are you getting fewer impressions in Google Search Console?

A falling line in Google Search Console feels like bad news. Even so, there is no need to panic straight away. Impressions are a visibility signal, not your revenue. A drop can be real, but in 2026 it is just as often a measurement change, a shift in the results page or a changing mix of search queries.

So always look at three layers at once: impressions, clicks and conversions from your analytics or CRM. Are your impressions falling while clicks and conversions stay flat? Then something has probably changed in the measurement, not in your performance.

Google normally shows around ten results per page. One parameter made it possible to request a hundred or more results in one go. SEO tools used that to pull an entire top 100 in a single request. All of those requested results counted as impressions, including for pages no real searcher ever saw.

That option has not worked since the middle of September 2025. Google never announced it; what Google told Search Engine Land about it is that using that URL parameter is not something Google formally supports (Search Engine Land, with Google's response, checked on 25 August 2026). Since then many accounts show the same pattern:

• A sudden drop in impressions without a comparable drop in clicks
• Fewer visible search queries in your reports
• An improved average position while you have changed nothing
• A stronger effect on desktop, in practically every sector

What the num 100 effect does not mean: no falling rankings, no penalty from Google, no problem with the quality of your content and no real loss of visibility.

You prove the num 100 effect with one simple segmentation in Search Console. Split your data into two position groups:

• Positions 1 to 10
• Positions 11 to 100

Use the filters in the performance report and compare both groups before and after the drop. Is the drop almost entirely in positions 11 and higher? Then the measurement change is almost certainly the explanation. Your visibility with real searchers, in the top ten, is simply intact. Is the drop in the top ten as well, then something more is going on and you work through the causes below.

1. Space on the results page and features

Google keeps filling the results page with AI summaries, shopping blocks, maps and forums. That leaves less room for the classic blue links. You can hold the same position and still appear less often. You recognise it by steady positions with falling impressions and slightly falling clicks, mostly on informational queries. Strengthen your FAQ sections, improve your snippets and switch on structured data.

2. Query mix and intent

Search traffic shifts from broad to transactional queries, or from brand queries to non-brand queries. Your revenue can even improve while your impressions fall. Quick test: label your top 100 queries as branded or non-branded and steer on that.

3. Indexing and canonicals

Google consolidates variants aggressively and ignores thin pages. If Google indexes different URLs from the ones you promote, your visibility can splinter. Symptoms: unexpected impressions on the wrong URLs, canonical conflicts and a growing pile of near duplicates. Merge variants, keep your internal links consistent, repair canonical tags and prune surplus pages out of the index. If everything looks technically fine but your site stays invisible, also read why your WordPress site is not in Google while everything looks fine.

4. Content quality and cannibalisation

Several pages aiming at the same intent force Google to choose. More pages is not by definition more SEO. More pages is often more noise.

5. Technology and rendering

Slow servers, blocking scripts, differences in client side rendering, blocked resources, a wrongly set robots instruction or the aftermath of a migration. These problems usually arrive gradually, not with one big bang. Check your coverage and indexing status, your sitemap and internal links, your redirects and canonical chains, and the status of your structured data and rich results. Speed plays a part here as well: look at why you score low on PageSpeed if you have doubts about your load time.

6. Authority and competition

Competitors improve themselves with expert content, better link structures, real mentions, stronger product feeds, reviews and a clearer brand. Then you fall without doing anything wrong yourself. A core update is the moment that reshuffling becomes visible all at once: what the May 2026 one did to the local pack, and which four actions hold up afterwards, is in the Google May 2026 core update. If you want the whole build up instead of the repair, read how do I get my website to the top of Google.

7. Data and attribution

Consent flows, browser restrictions, tag choices and definitions in your analytics all create differences between systems. So record per report which definition you are using, otherwise you are comparing two numbers that do not measure the same thing.

Work through these six steps before you change anything at all:

• Segment positions 1 to 10 against 11 and higher
• Segment desktop against mobile
• Compare your top pages on clicks before and after the drop
• Check the indexing report for falling coverage per template
• Look at the status of your rich results and schema warnings
• Work out whether the drop sits in branded or non-branded queries

After this checklist you know whether you are dealing with a measurement change, a shift on the results page, an indexing problem or a real content problem. Only then do you pick an action.

The biggest damage comes from reacting without a diagnosis. Avoid this:

• Rewriting content without a clear cause
• Creating new pages without a cannibalisation analysis
• Restructuring your internal links wholesale without a strategy
• Making technical changes without a measurement before and after
• Trusting new tools blindly
• Carrying year on year comparisons through while the basis of measurement has changed

Almost every wrong conclusion about falling impressions is a comparison problem rather than a visibility problem. The graph in Search Console is honest, but the two periods you lay against each other rarely are. Three traps catch people out again and again, and all three are avoidable in a minute.

Trap one: unequal periods

Search Console offers handy buttons for the last seven, twenty eight or ninety days. The trouble starts when you compare a period that contains five Mondays with one that contains four, or a stretch that holds a public holiday with one that does not. Almost every business has a weekday rhythm, and searching is no exception. Compare in blocks of twenty eight days, always ending on the same day of the week, and a large part of the noise disappears before you have interpreted anything.

Trap two: the last day that is not finished

Search Console data arrives with a delay, and the most recent days are not complete. If your comparison period runs up to yesterday, the tail of your graph is always lower than reality, and that dip is not a signal. Leave the last three days out of any comparison you intend to draw a conclusion from. This one detail explains a surprising number of panicked messages that arrive on a Monday morning.

Trap three: the mixed property

A domain property collects everything: your main site, your subdomains, your staging environment if it ever got verified, and any language versions you run. A change in one of those parts moves the total without anything happening to the part you actually care about. Filter by page path before you compare, or use a separate property per section. Running a multilingual site? Split by language folder first, otherwise a translation project shows up as a collapse.

None of these three is exciting, and that is exactly the point. Before you rewrite a single page, make sure that the two numbers you are comparing measure the same thing over the same kind of period. Most drops shrink to half their size the moment the comparison is fair, and the half that remains is the half worth investigating.

Suppose your split shows that the drop is not only in positions eleven and higher. Then real visibility has moved, and it helps to know where it tends to move to. The results page of 2026 is not the ten blue links it was, and the space above you is filled by things that answer the question before anyone scrolls.

• The answer sits above the links. On informational queries an AI summary often stands where your snippet used to be. The page below it still exists, and you can still be quoted in the summary itself, but the number of people who travel down to your entry falls. You recognise this pattern by informational queries losing ground while your commercial queries hold steady.
• People ask an assistant instead of a search box. A part of the questions your articles used to answer is now put to a chat assistant directly. That traffic does not appear in Search Console at all, because it never was a Google search. It is not lost demand, it is demand you can no longer see in this particular report. What you can do about it is a separate discipline, and it is covered under AI visibility.
• Your own brand absorbs the shift. When people already know you, they search for your name and skip the generic query altogether. That looks like a loss on non branded terms and a gain on branded ones. It is usually good news wearing a bad disguise, and you only see it if you label your queries before you judge them.
• A competitor took the block, not the position. Featured blocks, product panels and map results are winner takes most. Losing one of those costs you a large share of impressions on a query where your ranking position did not change at all. Structured data and a complete profile are what get you back in, not more words on the page.

The practical consequence is that impressions on their own are becoming a weaker measure of how well you are doing. Clicks are better. Enquiries are better still. If you have to pick one number to watch every month, watch the number of people who contacted you, and use Search Console to explain the movement rather than to define it. The longer arc of that shift, and what still works inside it, is set out in how to create content for AI assistants.

The checklist gives you one of four outcomes. Each of them has its own sensible next move, and the most common mistake is applying the response that belongs to a different outcome. Rewriting content because of a measurement change is the classic example: weeks of work against a problem that was never there.

Outcome one: a measurement change

The drop sits almost entirely in positions eleven and higher, clicks are flat, conversions are flat. Response: change nothing, and change your reporting instead. Add a note to your dashboard on the date the drop began, so that everyone who looks at the graph in six months knows what the step is. Then switch your monthly reporting to clicks and enquiries as the headline numbers, with impressions as context underneath.

Outcome two: the results page changed around you

Positions steady, impressions down, clicks slightly down, mostly on informational queries. Response: work on being the answer rather than on being longer. Put the actual question in the heading, answer it in the first two sentences, and make sure your structured data is valid. Then check which of your pages still earn clicks and strengthen those first.

Outcome three: something technical is holding you down

A whole page type loses coverage, or unexpected URLs appear in your reports. Response: stop the content work entirely until this is fixed, because you cannot judge content that is not being served properly. Work through indexing, canonicals, redirects and rendering in that order, and check the speed side as well at why you score low on PageSpeed.

Outcome four: your content genuinely lost ground

Clicks and conversions fall along with impressions, on queries that used to be yours, and nothing technical explains it. Response: this is the only outcome where rewriting is the right answer, and even then not everywhere at once. Take the five pages that lost the most clicks, establish per page what the searcher actually wants now, and rebuild those five before touching anything else. The full sequence for that is in how do I get my website to the top of Google.

Write down which outcome you landed on and on what evidence. Three months from now, when the graph moves again, that note is the difference between a diagnosis and a guess.

Does fewer impressions mean a lower ranking?

Not by definition. Because of the num 100 effect your impressions can fall sharply while your positions with real searchers stay the same or even improve.

Why is my average position improving while my impressions fall?

That is an arithmetic effect. When the impressions on positions 11 and higher fall away, mostly your good positions are left and the average rises by itself.

My clicks are falling too, what now?

Then more is going on than a measurement change. Think of broader changes on the results page or a shifted search intent, and work through the seven causes above.

Can there be technical problems without errors in Search Console?

Yes. Rendering, internal links and canonicalisation can hold your visibility down without an error message appearing anywhere.

What is the best way to compare my figures?

Work with windows of 28 days. That filters weekday effects out and compares like periods with like.

What is the first action with the most impact?

Strengthen your top 20 pages. That is where your proven visibility sits and where every improvement pays off soonest.

Do you want to learn more about Search Console, indexing and visibility yourself? Our knowledge base holds more explanation and practical step by step plans. Stuck halfway, then we are happy to sit down and look at the graph with you. In a quick diagnosis session we establish with segmentation and evidence where the drop comes from: a measurement change, a change on the results page, an indexing problem or a content problem. After that you get a prioritised action plan. Look at what we do for you when you want to improve your SEO.

If you would rather have a second pair of eyes on the graph before you decide anything, send us the two screenshots the split in FIG.01 produces, together with the date the drop started and the kind of business you run. We will tell you which of the four outcomes above yours is, and we will say so plainly when the honest answer is that nothing is wrong and you should leave your site alone. You can arrange that through the contact page, where you can pick a slot in the diary yourself, and there is no charge for a first look.