You are already ranking for more than you get clicked for.Win the clicks you have already earned.
On-page SEO that targets the one thing a searcher actually reads before deciding whether to click: the title and description Google shows them.
A page can sit at position 7 and earn almost nothing because its title was written for the catalog rather than the customer. That demand is already paid for, and it is being wasted.
The objective has a name: CTR optimization, raising the share of people who see you in the results and choose to click. Page title optimization and metadata optimization are the two levers inside it, because the title and the meta description are the whole of what a searcher reads before deciding whether you are worth a click. On most sites nobody has revisited either since the page was built.
A worked example, and it was one word
A client category page ranked between 1.7 and 4.1 for every variant of its head term, including the two phrases its buyers actually type. Its title contained neither word. It was the largest page in its batch, and the fix was a title.
Largest single page in one batch, before and after
Thirteen clicks to fifty-nine, click-through rate from 1.00% to 2.58%, on a page that had been ranking the whole time. Nothing about the page’s authority changed. Only what the searcher saw.
One objective, two levers
Page title optimization
The line a searcher reads first and the strongest single signal of whether the page answers their question. Rewritten to the character budget your CMS actually renders, in the words buyers use rather than the words your catalog uses.
Metadata optimization
The description underneath it, plus the keyword field where your platform still uses one. This is where the promise gets made specific enough to be worth acting on.
Both sit inside CTR optimization, which is the objective rather than a separate task. Neither one is judged on its own: the only question that matters is whether more of the people already seeing you decide to click.
How a batch runs
1. Find the demand you already have
Pages with impressions and no clicks. High position, low click-through rate. That traffic is already earned and it is leaking. We share the list of twenty before touching anything.
2. Freeze a baseline, then rewrite
Performance is recorded before anything changes. Then the title, meta description and keywords are rewritten to the character budget your CMS actually renders, for the searcher rather than the catalog.
3. Verify live, field by field
Every page checked after applying, not assumed. Submitted to Google and Bing, with the submission quota delta as proof it registered rather than a claim that it did.
4. Read it at 14 and 28 days
Against a control group of pages we deliberately did not touch. Directional at 14, decision at 28, and labelled as such in the report.
What you will see, and when
Baseline frozen, twenty URLs agreed, falsifier recorded. Nothing is applied yet.
Directional. Enough signal to steer, not enough to conclude, and the report says so.
Decision. Four clean weeks removes day-of-week bias. Continue, change, or stop.
Impressions are not the goal. Clicks are.
It is useless to generate impressions that never turn into visits. When a page ranks and nobody clicks, the problem is almost always the title and the meta description, not the content underneath them and not the authority of the page.
That is why this work is measured on organic CTR rather than on rankings. A page that moves from position 9 to position 7 and still earns nothing has not been improved. A page that stays at position 7 and doubles its click-through rate has.
The same sentence holds in AI search, with two words changed.
Search engines
Impressions are not the goal. Clicks are.
You were shown to someone. Whether they chose you is a different question, and it is decided by the title and description they read.
AI search
Inclusions are not the goal. Citations are.
Your page was in the model’s candidate set. Whether it was quoted, and whether your name survived into the answer, are two further questions again.
Being present is not the same as being chosen, in either channel. Both halves are worth buying and neither can be judged by the other’s numbers, which is why we sell and report them separately.
What makes this measurable rather than plausible
Three rules. They are the reason to choose this over a rewrite you could get anywhere, and they are easy to check.
A control group, not a site average. We compare against pages we chose not to touch. Otherwise we are measuring our own other work, the season, or the algorithm.
The concentration test. Remove the batch’s biggest page and re-run the numbers. If the result was one page wearing a batch’s clothes, you will be told. That test has killed one of our own headline numbers.
A falsifier, recorded before we start. We write down in advance what result would mean the work did not succeed. If we cannot name it, it is an opinion and we label it as one.
The evidence, including the part that went backwards
117 pages across six batches, June to September 2026, on a single client. That account is an unusually clean test environment: no paid search, no shopping campaigns, no ads budget. Every result is organic and uncontaminated by spend.
Click-through rate change, nine weeks after applying
The best measured batch moved 63% against a control group at 32.7%, and it survives the concentration test.
One batch, twenty pages
And on that batch, eight of twenty pages declined. Both numbers are in the report, because a program that only ever reports wins is not measuring anything.
What this instrument cannot tell you. Google withholds low-volume search terms, so some movement is always unattributable. Click-through rate is affected by how many results Google chooses to show and by AI Overviews appearing above them, neither of which we control. We report the number and the limit together.
Where this works, and where we will say no
It fits large catalogs, templated or generic titles, specific and measurable demand, and pages that can convert once someone arrives.
We will decline it where:
- There are fewer than roughly sixty qualifying pages. The program exhausts itself in two batches and you should not pay for a third.
- The local pack carries the traffic rather than organic results. The clicks go to your Business Profile, and organic click-through rate is the wrong scorecard entirely.
- Nothing is tracked as a conversion, so the third measure cannot be reported and you would be judging the work on two thirds of the picture.
- No baseline can be frozen before work starts. Without it, nothing afterwards is provable.
Where it sits beside the rest of the work
This is one half of the on-page program. The other half is answer engine optimization, which structures the same pages so AI assistants can quote them. They act on different channels and are measured with different instruments, so we sell and report them separately.
There is a third step in the AI channel that most reporting misses entirely. A page can be included, then cited, and still have the brand go unnamed in the answer. We have measured exactly that on a live account: the source was used, the company was not credited. Inclusion, citation and attribution are three different questions, and only the third one sells anything.
Underneath both sits technical SEO, because a page that cannot be crawled, indexed or rendered will not benefit from a better title. And the wider picture is on the SEO consultant page.
Questions we get asked
Q. What exactly gets changed on the page?
The title tag, the meta description and the meta keywords, written to the character budget your CMS actually renders rather than a generic best practice number. Body content, design and navigation are untouched. That is deliberate: changing one layer is what makes the result attributable. If we changed five things at once, neither of us would know which one worked.
Q. Why only twenty pages at a time?
Because twenty is large enough to read statistically and small enough to attribute. It also forces prioritisation: the pages carrying the most wasted demand get done first. You approve the twenty before any work starts, so nothing is a surprise.
Q. How many batches will we need?
Expect three or four productive ones, not an indefinite retainer. The head pages carry most of the value and they get done first by design, so the return falls off. We will tell you when the pool is exhausted rather than keep billing. If your site has fewer than about sixty pages worth optimizing, the program exhausts itself in two batches and we will say so before you buy.
Q. What if impressions fall after the rewrite?
Falling impressions are often the intervention working. A more specific title matches fewer queries, and the ones it does match are closer to what you sell. That is why we report impressions, clicks and conversions as three separate questions rather than rolling them into one score. A page that loses impressions and gains clicks has done exactly what we asked of it.
Q. How do you prove it was your work and not the season?
A control group. We pick pages we deliberately do not touch and compare against them, so seasonality, algorithm updates and your other marketing affect both groups equally. On our best measured batch the pages we rewrote moved 63% against a control at 32.7%, over the same nine weeks.
Q. What happens to the pages that get worse?
They go into a rework register with the size of the prize stated honestly, and they appear in the report alongside the wins. On one batch, 12 of 20 pages improved and 8 declined. You are reading that here rather than discovering it in month three.
Q. You call it on-page SEO. Our agency calls it CTR optimization. Same thing?
Close, and the difference is worth getting right. CTR optimization is the objective: raising the percentage of people who see you and click. Page title optimization and metadata optimization are the two levers that deliver it, so they sit inside CTR optimization rather than beside it. On-page SEO is the wider category that contains all of it, and it is the term most buyers search with, which is why it is the name on this page. What matters more than the label is whether anyone is measuring the outcome against a control group. Most of the time nobody is.
Q. Is this the same as AI search optimization?
No, and the distinction matters commercially. This work targets the search results page: the words a person reads before deciding to click. AI visibility work targets whether an answer engine can quote you at all, and it is measured with completely different instruments on a slower timeline. They are sold and measured separately for that reason. Judging one by the other’s numbers is the most common way both get written off.
Q. Will you work with our existing agency or team?
Yes, frequently. This is a bounded piece of work with its own measurement, so it sits alongside whatever else is running without disrupting it. Your team sees the list, the baseline, the rewrites and the reads, which means they can run the next batch themselves if they want to.
Q. What do you need from us to start?
Search Console access, the ability to edit titles and meta descriptions in your CMS, and someone on your side who can approve a list of twenty URLs. If conversions are tracked we will use them as the third measure. If nothing is tracked, we will tell you that before quoting rather than after.
Start with a conversation
Thirty minutes to look at whether you have the kind of wasted demand this fixes. We can usually tell from Search Console within the call. If your pages are not losing clicks this way, we will say so rather than sell you a batch.