A rising impression count can leave you wondering why the website still receives so few visits. Start with one page and the searches that brought it into the results, because the sitewide total can hide whether people are seeing a relevant service, an unrelated answer or a result far below the choices they consider.
Check that distinction before making another round of changes across the website. Sometimes Google shows a page to people searching for something else entirely, and sometimes the page matches the search but people still pick another result. The two need different fixes, and you have to look at the searches themselves to know which one you have.
This guide gives you a way to choose that work. If the website has only just launched, pair it with the first 90 days measurement plan, so you judge a page once it has had time to find an audience.
First, separate three questions
Search Console tells you about search visibility and clicks. Your website analytics tells you about observed visits and actions, and your customer records tell you whether an inquiry became useful business. Keep those questions separate while you investigate.
Google defines click-through rate as clicks divided by impressions. Its average position describes the topmost result from your property, averaged across impressions. Read both numbers page by page, since a sitewide average mixes pages and searches that have little in common, and look for real buyers in your inquiries.
For example, imagine a site has 900 impressions for a broad research article and 100 for a local service page. If the article is rarely chosen, it can dominate the overall click-through rate even when the service page is doing useful work. Improving the article may still make sense, but that decision should depend on who it serves.
Start with one business question: which service or problem do you want the next customer to find? Write it above your spreadsheet. It gives you a reason to choose one set of impressions over another.
Make a small report you can use
Open the Search results performance report. Select a complete period, such as the last 28 days, and use the same search type, country and device filters when you compare it with another period. Leave the most recent incomplete data out of close comparisons.
Choose the page you care about in the Pages tab, then open its Queries tab and export the useful rows. Keep separate notes for searches that contain your business name and searches from people who may never have heard of you.
Use this working table:
| Field | What to write |
|---|---|
| Page | The exact URL you want to improve |
| Search group | Closely related queries with the same practical need |
| Audience fit | Customer, researcher, job seeker, or unclear |
| Impressions and clicks | Counts for the same filters and period |
| Average position | Context for visibility, which varies from search to search |
| Page promise | What the title and opening offer the reader |
| One proposed change | A change tied to the diagnosed problem |
| Review date | When you will compare equivalent data |
Do not expect the exported query rows to add up to the chart. Search Console uses different aggregation rules in different views and leaves some queries out, so treat the report as a decision aid with gaps in it.
Follow the branch that matches the evidence
The searches do not match the business
If a commercial cleaning company appears for searches about cleaning a home oven, those impressions may have little sales value. Changing a description to win more of those clicks would reward the wrong outcome.
Check why the page attracts that search. A broad title or a long digression can confuse who the page is for, so make the intended audience clear in the title and opening, and carry it through the examples and internal links. Keep useful background material if it helps the right reader, and judge it by a different target from the inquiries you expect of a service page.
Some educational pages serve future customers well. The test is whether you can explain their connection to a real buying decision, and “It gets impressions” is an incomplete reason to expand a topic.
The searches fit, but visibility is weak
A relevant page that appears mostly well below the leading results has a visibility problem before it has a title problem. Review whether it answers the question people are asking with enough substance to help them.
For a service page, that may mean the work offered, location served, practical constraints, process, evidence and next step. For a guide, it may mean a worked example or decision tool that competing summaries do not provide. Google’s content guidance emphasizes original value and useful, complete answers, so add length only when it adds help.
Also check whether other relevant pages on your own site link to this one. A useful service explanation hidden from the main navigation and related guides is harder for readers to discover. Choose the most obvious improvements first, then give them time to collect observations.
The searches fit and the result is being seen, but few choose it
Now inspect the result itself. Search from the relevant location and device when you can, and treat that view as a sample, since your own search shows one version of the results.
Compare the page promise with the other choices on the screen:
- A search for a practical troubleshooting guide should lead to a title that sounds like a guide and does not read like a product pitch.
- A local service page should clearly name the service and the area.
- The description should say something concrete that could only belong to your business.
Write two candidate titles on paper and choose the one that most accurately describes what the page helps with. Update the page opening so it delivers on that title. Avoid promising a checklist when the page holds only an opinion, or naming a city you do not serve.
Search results can also answer a question directly or show maps and videos, which gives some searchers less reason to click through. A more persuasive title cannot create a need for a visit that the searcher does not have.
The page barely has any observations
With very small counts, one click changes the rate a great deal. For example, one click from 20 impressions is 5%, and zero clicks is 0%. Compare versions of a page over the same searches and dates before you call one better.
Keep collecting comparable data while you fix issues you can see directly, such as an inaccurate title or an unclear offer. We would avoid weekly title changes driven by swings in the percentage, because they take away the stable period of evidence you need to connect an edit with a result.
Check an unexpected absence separately
If a page has no impressions at all, inspect the URL before treating the situation as a low-click problem. Google’s URL Inspection tool separates information about the indexed version from a live test. A successful live test does not promise that the page will appear in results.
Confirm you are examining the intended URL, including redirects and any canonical version Google has selected. A developer can then look for an accidental noindex instruction or blocked access, even an incorrect canonical. Fixing such an issue makes the page eligible to appear again, while demand and position still depend on the searches people make.
Run one useful improvement cycle
Choose a page with clear customer relevance and enough observations to learn from. Record the diagnosis, make one coherent set of changes and note the date. A title and opening that need to agree can change together, while rewriting the whole site at once makes the cause of any change much harder to see.
At the review date, check whether the relevant query group appears more often and whether people are choosing the page, then whether those visits produce useful inquiries. Compare the country and device mix before you credit every change to your edit. Seasonality and competitors can change too, and so can the search results themselves.
If clicks improve but inquiries do not, take the investigation further down the path. Check the offer and the contact path, and confirm that submissions arrive where staff work. The website-to-CRM tracking example shows how to prove that handoff.
A useful outcome from this exercise can be modest, such as one page to improve and one irrelevant topic to stop worrying about, even a broken contact path to repair. In our view that gives the next month of work a firmer basis than a larger impression total with no explanation behind it.