You open Search Console and see impressions rising while clicks fall. Publishing more content may explain the increase in impressions, but the drop in visits is harder to account for. Changing titles immediately or concluding that AI search is to blame can lead you to fix the wrong thing.
First, check whether the increase in impressions and the decrease in clicks came from the same queries and pages. Before judging overall CTR, I examine where clicks fell and where new impressions grew separately. The two changes may have happened in different places.
1. Align the comparison periods and search type
CTR is clicks divided by impressions. If there are 10,000 impressions and 500 clicks, CTR is 5%. So if impressions rise and clicks fall, overall CTR has indeed declined. That calculation alone, however, does not establish a problem with content quality or titles. See Google's Search performance report guide for metric definitions and comparison settings.
Set the date filter to Compare in Search Console's Performance report and compare the same search type. Mixing image search data into an analysis of web search clicks means analyzing a different phenomenon. If the current month has not ended, do not compare it with the whole previous month; select completed periods of equal length. Comparing the last 28 days with the preceding 28 days is one way to keep the weekday mix consistent.
For highly seasonal products, also examine the comparable period in the previous year. This helps distinguish a month-over-month decline from a seasonally expected decrease. On the first line of your analysis notes, record the periods, search type, and country and device filters. A colleague should be able to open the report later and see the same numbers.
2. A change in the query mix also changes overall CTR
The following is hypothetical data created to explain the principle. It is not a measurement from an actual site. The overall result can change even when click-through rates remain unchanged for searches by people who already know the brand and for broad informational searches.
| Query group | Impressions: before → after | Clicks: before → after | CTR by group |
|---|---|---|---|
| Queries containing the brand name | 5,000 → 2,000 | 800 → 320 | 16% → 16% |
| General informational queries | 5,000 → 18,000 | 50 → 180 | 1% → 1% |
| Total | 10,000 → 20,000 | 850 → 500 | 8.5% → 2.5% |
Neither group's CTR deteriorated. Overall CTR fell because impressions for branded searches, which generated many clicks, decreased while impressions for low-CTR informational searches grew substantially. In this situation, changing every article title is not the priority. Investigate why brand demand fell and whether the new informational queries provide visibility the business needs, as separate questions.
A useful starting point for an actual analysis is to separate brand names, product names, and general exploratory queries. Account for variations in brand spelling and spacing, but do not assume the query table includes every search. Privacy-related anonymization and aggregation methods can cause the table totals to differ from the chart. Do not fill the gap with invented queries to make them match. Search Console's guide to reporting discrepancies explains these differences in more detail.
Also read the queries that have newly increased. Being found in a product comparison just before purchase serves a different purpose from being found for a broad definition. More visibility during information gathering may be a new opportunity, but it does not explain away lost clicks from existing purchase-related queries. This does not mean it is acceptable to lose any particular query; it means traffic serving different purposes should not all be judged by one click-through rate.
3. Find the pages contributing most to the decline in clicks
Start with the whole site and inspect the pages with the largest losses in clicks. A page falling from 10 clicks to 2 has an 80% decline, but a page falling from 1,000 to 800 has lost far more clicks. Record both the percentage decline and the absolute loss when deciding which problem to investigate first.
After selecting a page, examine its queries. Separate queries that lost clicks from those that gained impressions, then break them down by country and device. If the decline occurred only on mobile in Korea, for example, an improvement in the overall average caused by more desktop impressions overseas is a separate matter. Google's guide to analyzing search traffic drops also recommends checking differences by page, query, country, and device.
If several important pages dropped sharply on the same day, check deployment history and indexing status before editing titles. If the decline is confined to queries for a seasonal product, changes in demand for those searches come first. Google Trends can provide context about the direction of interest, but its index cannot be converted into the number of clicks your site lost.
It is also worth checking whether another page on your site has started appearing for the same query instead of the original article. Looking only at the old page's clicks may suggest a large decline even though visits moved to the new page. In that case, check whether both URLs answer the same question appropriately and whether visitors can still find the information they need after the shift. If an unrelated page is appearing instead, review the content and the role of each page.
4. The same average position does not mean the same placement
To answer “Why did clicks fall when the average position is still 5?”, first examine what makes up that average. Yesterday's position 5 and today's position 5 do not necessarily represent the same queries, devices, or people. The average can remain similar even if positions rise for some queries and fall for important ones.
Position in Search Console is a relative position aggregated across search impressions. The counting method also varies by search result type. Do not read the overall average as a fixed rank. The precise definitions appear in the guide to counting impressions, positions, and clicks.
Narrow the comparison to the same page, query, and device as far as possible. Even then, a position number cannot fully describe screen placement or the likelihood of a click. For queries with small samples, do not interpret minor movement over a few days as a long-term decline; check whether it persists over a longer period.
5. Check titles and descriptions in the actual search results
If impressions and position for the same query are broadly stable but clicks have fallen, examine what readers currently see in the results. Look for ads, product sections, maps, videos, and summaries, and read whether the title and description alone make clear what your page offers.
For example, if a page is titled “The Perfect Way to Improve Work Efficiency” but actually provides a meeting notes template for small teams, readers may struggle to recognize its value before clicking. You could consider an edit that reflects the content, such as “Meeting Notes Template for Small Teams: Record Decisions and Owners.” This is an illustrative title, and any increase in CTR after the change is unknown until measured.
The title you write does not necessarily match the title shown in search. Google uses several sources to generate title links. Descriptions are not fixed to the meta description either; Google may select relevant body text for a snippet based on the query. Check that the title, description, and body make the same promise.
6. AI summaries are a possibility to investigate, not an automatic explanation
Google's guide to AI features in Search explains how AI Overviews and AI Mode summarize information and provide links to relevant web pages. How much these features contributed to your site's decline in clicks is a separate question. You cannot calculate that share from the overall impressions and clicks graphs alone.
The presence of an AI summary in the current results is not proof that it was absent before. If you did not save the earlier results, record it as “currently observed.” Also note the query, country, device, observation date, and whether you were signed in, so later observations can use comparable conditions. Observing the results helps narrow possible causes, but it does not replace data on all users' actual clicking behavior.
When the team disagrees, distinguish “AI has an effect” from “AI explains most of the decline.” The former may be a hypothesis you can investigate in the search results, while the latter requires evidence that also rules out other changes. If query mix, brand demand, and page edits changed at the same time, it is more accurate to state that the causes have not yet been separated. You can still proceed with fixing verifiable title errors or outdated information.
7. Record the established facts and next action, one line at a time
At the end of the analysis, state the scope rather than simply writing “CTR fell.” For example: “Over the last 28 days, Page A lost clicks from product comparison queries on mobile in Korea. Its title and body still mention a model no longer sold, so we will update them and check the same query group again.” This defines the next task without dressing up the result.
Record the comparison conditions, affected pages and queries, established facts, remaining unknowns, and the changes made with their dates. If you change titles, body content, and site structure all at once, it becomes difficult to separate their effects. Fix factual errors and access problems promptly, but give improvement hypotheses a defined scope wherever possible and observe them. These records make both understanding the growth in overall impressions and recovering lost clicks more concrete.