Three hundred and four thousand times, Google decided this site deserved to be seen. Six hundred and forty-eight times, somebody clicked.
Publishing that is uncomfortable, which is precisely why it is worth doing. Most articles about click-through rate use a hypothetical example or an anonymised client. This one uses our own Search Console data, because the gap it reveals is the single most common and most fixable problem in ecommerce SEO, and because a real number is more useful than an illustrative one.
The finding is not that our content is bad. It ranks — 304,000 impressions is Google repeatedly deciding these pages belong on the results page. The finding is that the twelve words that represent each page in the search result were doing almost none of the work they exist to do.
This is the diagnosis, the benchmarks, the rewrite system, and the workflow to do it at scale.
CTR gap — the difference between the click-through rate a page achieves and the benchmark rate for its average ranking position. A meaningful negative gap indicates that the page is being shown and rejected, which is a title and description problem rather than a ranking or content problem, and is therefore fixable without earning a single additional link.
Our Own Numbers, In Full
| Metric | evolveamz.com, 6 months | What it should be |
|---|---|---|
| Impressions | 304,000 | — |
| Clicks | 648 | 3,000–4,500 |
| CTR | 0.21% | 1.0–1.5% |
| Average position | 11.9 | — |
| Gap | Roughly 2,400 to 3,900 clicks, no ranking change required | |
What the numbers say
An average position of 11.9 is the top of page two — visible, competitive, and genuinely earned. At that position a clean search result typically returns somewhere between 1.0% and 1.5% CTR. We were at 0.21%, which is between five and seven times below where the same rankings should land.
The pages carrying the most impressions were the ones with the widest gaps. Three in particular — a local business AI search guide at roughly 23,000 impressions, an AI crawler list at 21,800, and an influencer whitelisting guide at 15,300 — were generating substantial visibility and almost no traffic.
Three hundred thousand impressions is Google repeatedly deciding you belong on the page. Six hundred clicks is searchers repeatedly deciding you do not.
Why this is good news
A ranking problem takes months and links to solve. A click-through problem takes an afternoon and a text editor. The impressions already exist — the audience has already been assembled and shown to you. Closing even half this gap is worth more than most link-building campaigns, and it is entirely within your control.
Finding Your Low-CTR Pages in GSC
Exact steps. Fifteen minutes.
- Open Search Console → Performance → Search results.
- Set the date range to the last six months. Shorter windows are noisy; longer ones include stale content.
- Enable all four metrics — clicks, impressions, CTR and average position. The default view hides the two you need.
- Switch to the Pages tab and export to a spreadsheet.
- Filter to rows above 1,000 impressions. Below that, single-query variance swamps the signal.
- Add a column for expected CTR using the benchmark table in the next section, matched to each row's average position.
- Add a gap column: expected CTR minus actual CTR, multiplied by impressions. That gives missed clicks per page.
- Sort by missed clicks descending. That is your work queue, in priority order.
Then do it again at query level
Repeat on the Queries tab. Page-level data tells you which pages to fix; query-level data tells you what the searcher actually wanted, which is what the new title has to answer. A page ranking for a query you did not write it for is extremely common and is often the entire explanation for a low CTR.
Sort by impressions descending, then look only at rows where average position is under 15 and CTR is under 1%. Those are pages Google already trusts enough to show prominently, being rejected at the moment of choice. They are the cheapest clicks available anywhere in your marketing.
Expected CTR by Position in 2026
| Position | Clean SERP CTR | With AI Overview present |
|---|---|---|
| 1 | ~27% | Substantially lower; studies show 30–58% reduction |
| 2 | ~15% | Reduced proportionally |
| 3 | ~11% | Reduced proportionally |
| 4–5 | ~6–8% | Reduced proportionally |
| 6–10 | ~2–4% | Reduced proportionally |
| 10 | ~1.7% | Often under 1% |
| 11–15 | ~1.0–1.5% | Often under 0.7% |
Important caveats on this table
These are clean-SERP figures, meaning a results page with no local pack, shopping carousel, video block or AI Overview. In 2026 that is an increasingly rare page. Treat the table as a base case for planning and always compare it against your own Search Console data for diagnosis, because your actual SERPs will have features these numbers do not account for.
Device matters too: mobile CTR runs meaningfully below desktop, with position one at roughly 26.9% versus 32%. Most searches are mobile, so if your titles truncate badly on a phone you are losing where the volume is.
How to use it honestly
Do not treat a below-benchmark CTR as automatic proof of a bad title. Check whether an AI Overview or another SERP feature is present on your main queries first, because that changes what "expected" means. The diagnosis in section five separates those cases.
Why the Baseline Moved
If you are comparing against benchmarks you learned before 2024, you are measuring against a world that no longer exists.
- Position one CTR fell from about 39.8% in 2022 to roughly 27.6% in 2026 — a 30% decline in four years. The trend has been monotonic, with the sharpest single year being 2024 as AI Overviews rolled out.
- AI Overviews now appear on a substantial share of results pages, with estimates ranging from roughly 13% to 31% depending on the dataset and query mix, up from a negligible base in mid-2024.
- Where an AI Overview appears, organic CTR drops sharply. One widely cited analysis found overall organic CTR falling from 1.76% to 0.61% on affected queries, a decline of around 65%.
- Queries without AI Overviews have also declined, which suggests behavioural change rather than a single-feature effect. Users have learned to expect answers rather than links.
- Forward modelling suggests position one settles in the 22 to 25% range by 2027.
The case that shows the ceiling
A major publisher's page ranking at position one for a celebrity query saw daily clicks fall from around 6,000 to roughly 100 after Google introduced an AI Overview for that query — with no change in ranking whatsoever. The Overview answered the question completely, leaving no reason to click.
That is the extreme case, on a purely informational query where the answer is a single fact. It is not representative of commercial or comparison queries. But it establishes what the mechanism can do when the query has one short answer.
Being cited inside an AI Overview partially offsets the loss — cited sites see meaningfully better CTR than uncited ones on the same queries. Which means title rewriting and AI citation work are complementary rather than alternatives. Our AI Overviews guide for ecommerce covers the citation half.
The Four Diagnoses
Title and description are not compelling. This is the rewrite case, and the one this article is about. Cheapest fix available.
A SERP feature is taking the clicks — featured snippet, People Also Ask, or an AI Overview. Rewriting helps less than winning that feature does.
Your title is working. Do not touch it. Spend the effort on moving the ranking up instead.
Impressions come from queries the page does not answer. No title fixes this; the content or the targeting has to change.
How to tell pattern one from pattern two
Search your top three queries manually, in an incognito window. If an AI Overview or featured snippet occupies the top of the page, you are in pattern two and the honest expectation for a rewrite is modest. If the results page is mostly blue links and yours is being passed over, you are in pattern one and a rewrite is the highest-return work available.
Our own case was predominantly pattern one, which is why the rewrite was worth prioritising over anything else on the roadmap.
Five Title Patterns That Outperform
1. Number plus specific outcome
"7 Amazon Listing Fixes That Raised Conversion 22%". Numbers set an expectation of scannable structure, and a specific outcome beats a vague benefit. Avoid round numbers where a real one exists.
2. Question matching the query exactly
"How Much Does Amazon Product Video Cost in 2026?". If the searcher typed a question, a title that is that question converts strongly. Pull the exact phrasing from your GSC query export rather than guessing.
3. Comparison with named entities
"Helium 10 vs Jungle Scout: Which Is Better for New Sellers?". Naming both options captures searchers researching either one, and comparison intent has high click intent because the AI Overview rarely resolves a genuine trade-off.
4. Contrarian or corrective
"Why Your A+ Content Is Not Improving Rankings". Works when it contradicts a common assumption and you can actually support it. Do not use it where the payoff does not deliver, since bounce is worse than a missed click.
5. Current year plus definitive framing
"The 2026 Amazon Fee Structure, Explained". Year signals freshness, which matters to searchers and to AI engines. But this commits you to an annual update — a title with last year's date actively signals staleness.
Primary keyword in the first 30 characters. One clear benefit or outcome. No brand name unless it is the differentiator. Nothing that reads like every other result on the page. The test is simple: put your title next to the nine other results for that query and ask whether it is the one you would click.
Character Limits and Truncation
Google truncates by pixel width, not character count, which is why simple character rules fail on some titles and not others.
| Element | Safe length | Notes |
|---|---|---|
| Title, desktop | ~55–60 chars | Pixel-limited; wide characters truncate earlier |
| Title, mobile | ~50–55 chars | Where most of your traffic is |
| Meta description | ~150–160 chars | Frequently rewritten by Google regardless |
| Front-loaded portion | First 30 chars | Must carry the keyword and the hook |
The practical rules
- Write for mobile first. If it survives 50 characters it survives everywhere.
- Front-load ruthlessly. Assume everything after character 55 may never be seen.
- Capital letters and wide characters cost more pixels than lowercase and narrow ones. An all-caps word can truncate a title that would otherwise fit.
- Preview before publishing. A SERP preview tool takes seconds and catches the case where your best word is the one that got cut.
- Drop the brand suffix on long titles. "| Evolve Media Agency" on a 58-character title spends your last visible characters on something the searcher does not need.
When Google Rewrites Your Title Anyway
It happens often, and understanding why turns it from an annoyance into a diagnostic.
Why Google substitutes a title
- Yours is too long and truncation would lose meaning.
- Yours is keyword-stuffed or reads as boilerplate across many pages.
- Yours does not match the query a particular searcher used, so Google picks text from your page that does.
- Your H1 or on-page heading is a better fit for the specific query.
- Yours is missing or duplicated across the site.
What to do about it
Google most often substitutes text drawn from your own page — usually the H1 or a prominent heading. So the fix is rarely to fight the rewrite; it is to make sure every candidate on the page is good. If your H1 is strong, a substitution costs you little.
Treat a persistent rewrite as information: Google is telling you it thinks something else on your page describes it better than your title does. That is worth investigating rather than overriding. Google's own title link documentation sets out the behaviour and the influences.
Want us to run this on your site?
We will pull your Search Console data, calculate your CTR gap by page, and hand you the priority queue — whether or not you work with us afterwards.
Book a Strategy Call →The Ecom Profit Box
Eleven playbooks on listings, conversion, images, and email. Built for operators, no fluff, no email sequence.
Grab It Free →What Meta Descriptions Are Actually For Now
Google rewrites meta descriptions frequently — more often than titles — which leads a lot of people to conclude they no longer matter. That is the wrong conclusion for two reasons.
Reason one: they still get used
When your description genuinely matches the query, Google often keeps it. The rewrites happen when it does not match, which means a well-targeted description survives more often than a generic one.
Reason two: they are read elsewhere
Your meta description is the text that appears when a page is shared on social platforms, sent in messaging apps, and pulled into some aggregators. It is doing work well beyond the search result, and that work does not get rewritten by anyone.
How to write one that survives
- Answer the query in the first sentence. Not a preamble about your company.
- Include the specific detail that differentiates — a number, a date, a named comparison.
- State what the reader gets, concretely. "Full pricing table" beats "learn more about pricing."
- Use the searcher's vocabulary, taken from your query export.
- Keep it around 150 to 160 characters and front-load the important half.
- Skip the call to action. "Click here to find out more" consumes characters and persuades nobody.
Google's snippet documentation covers what influences whether yours is used.
The Batch Workflow
Doing this one page at a time in an editor is how the project dies at page eleven.
The process
- Export your gap analysis from section two, sorted by missed clicks.
- Take the top 20 pages. Not the whole site. You want a measurable cohort and a control group.
- Pull the top three queries per page from the GSC query export, so each rewrite answers a real search rather than your assumption.
- Draft in a spreadsheet with columns for URL, current title, new title, character count, current description, new description. Reviewing forty titles side by side surfaces repetition you would never notice one at a time.
- Check every title against the mobile limit before it leaves the sheet.
- Deploy in one batch, on one day, and record the date.
- Leave the rest of the site untouched for 30 days so you have a control.
Deployment on WordPress
For a Yoast site, the SEO title and meta description live in post meta, which means bulk updates are scriptable rather than manual. The relevant fields are _yoast_wpseo_title and _yoast_wpseo_metadesc, and WP-CLI can set them from your spreadsheet in a single pass.
Back up post meta before any bulk operation and run it against a single post first. A scripted rewrite that goes wrong across 200 pages is considerably harder to undo than it was to deploy.
Measuring the Lift
Two things make this measurement harder than it looks: a reporting lag and a re-crawl lag.
The timeline
- Days 1–3: nothing. Google has not re-crawled most pages yet.
- Days 4–14: re-crawl and re-index. New titles begin appearing, unevenly.
- Days 14–30: the first honest read. Search Console data also lags by roughly two to three days, so a 30-day window gives you about 27 days of usable data.
- Days 30–60: the reliable comparison window.
What to compare
CTR at constant position, not clicks. Clicks move with seasonality, ranking changes and impression volume. CTR against the same average position isolates the variable you actually changed.
Compare your rewritten cohort against your untouched control pages over the same window. If both moved, something else caused it. If only the cohort moved, the rewrite worked.
The single most common way this project gets abandoned is checking after a week, seeing nothing, and concluding it failed. Most pages have not been re-crawled by then and the reporting lag hides part of what has. Set a calendar reminder for day 30 and leave it alone until then.
For ongoing tracking, our Looker Studio reporting dashboard guide covers building a CTR-gap view that updates itself.
When the Rewrite Does Not Work
Sometimes CTR does not move. Here is how to read that rather than repeating the exercise.
Diagnose in this order
- Check whether Google is using your title. Search the query and look. If it substituted something, the rewrite never shipped from the searcher's perspective.
- Check for SERP features. An AI Overview or featured snippet that appeared since your baseline changes the achievable ceiling, and no title overcomes it.
- Check query intent alignment. If impressions come from queries the page genuinely does not answer, the title is not the problem — the page is targeting the wrong thing.
- Check for a ranking change. CTR looks worse if you dropped positions during the window.
- Check whether the promise is credible. An overreaching title gets clicks and then bounces, which is worse than not being clicked at all.
The two situations where a rewrite genuinely cannot help
The query has a one-sentence answer. If an AI Overview resolves it completely, there is no reason to click regardless of your title. The response is to target queries with answers that require nuance, comparison or judgment — which is also where commercial intent concentrates.
You are ranking for the wrong thing. Impressions on queries your page does not serve are not an opportunity, they are noise. The fix is content or targeting, covered in our guides to topical authority clusters and programmatic SEO for ecommerce.
What we did next
We took our own top 20 pages by missed clicks, rewrote titles and descriptions against actual query data, deployed in a single batch, and left the remainder as a control. This article publishes the starting numbers deliberately, so the follow-up can be checked against them rather than asserted.
The Short Version
- Our own six-month data: 304,000 impressions, 648 clicks, 0.21% CTR at average position 11.9 — roughly five to seven times below benchmark, representing about 3,000 missed clicks with no ranking change needed.
- 2026 benchmarks are lower than you learned: position one is around 27%, down from 39.8% in 2022, and position ten sits near 1.7%.
- Where an AI Overview appears, organic CTR drops 30 to 58% depending on the study, with one analysis finding an overall fall from 1.76% to 0.61%.
- Four diagnoses: rewrite the title, win the SERP feature, leave it alone, or fix the targeting. Manually searching your top queries tells you which one you are in.
- Front-load the primary keyword into the first 30 characters and write for the mobile limit of roughly 50 to 55 characters, since Google truncates by pixel width rather than character count.
- Google substitutes titles using text from your own page, usually the H1, so making every candidate strong beats fighting the rewrite.
- Batch 20 pages, leave the rest as a control, and judge at day 30 — not day seven, when most pages have not been re-crawled and reporting still lags.
External Sources Cited in This Article
- Google Search Central — Title link best practices
- Google Search Central — Snippet and meta description guidance
- Google Search Console Help — Performance report metrics
- Google Search Central — AI features and your website
- evolveamz.com Google Search Console, six-month window — first-party data published in section one