Seen, not chosen.
A services business was appearing in more searches every month in the market it most wanted to win, and getting no more visits from them. The audit found it was not one problem. The pages that should sell ranked too low to be clicked, the offer was hard to read, and a handful of defects got in the way. Here is what we found, the fix for each, and what we have not yet proven.
Start with the buying journeyDecision guide
Do customers use AI to choose?
Use GEO when it follows buyer behaviour.
Status
Audit complete and engagement started. Nothing below has been implemented and measured yet, so there are no results on this page. They will be added when they exist, including any that are disappointing.
The client
A business that sells high-value, made-to-order services to other businesses, mostly to buyers outside its home country. It also has a lower-priced product, and pages aimed at a long list of sectors.
The question was simple. Search Console said they were being seen more and more in the market they wanted. Why was that not turning into visits or enquiries?
We are deliberately vague about who they are and what they sell. The engagement is current, and our default is anonymous unless a client approves the exact text.
What the data showed
Visibility was growing quickly. Clicks were not. Impressions rose sharply over the most recent month of data while clicks stayed flat. Much of the growth came from the market they wanted.
The sector pages, which carried a large share of those impressions, appeared on average far below the first page of results. At those positions a page is technically shown and practically never seen. So the growth was real, and it was growth in being listed, not in being chosen.
We are not publishing the figures. They belong to the client.
What the data did not show
Two things a less careful report might have claimed.
- A penalty. Rising impressions with flat clicks has an ordinary explanation: pages ranking too low to attract a click. Nothing in the data needed a penalty to explain it, so we said so.
- A best market. The impressions told us where they were being seen, not which searches came from people who might buy. The country with the most exposure is a candidate to research, not a confirmed target.
Findings and fixes
| What we found | The fix |
|---|---|
| The offer was hard to read. The main service and the lower-priced product sat side by side. A visitor looking for something cheap and ready-made is not the buyer for a made-to-order project, and the pages did not make that clear. | On each commercial page, state the engagement model: scope, indicative cost, external charges, who owns the work, support and next steps. Target the searches that match what the business actually wants to sell. |
| Sector pages had words but not proof. They had plenty of sector-specific content, so the problem was not thin or copied pages. They lacked evidence a buyer could check. | Improve a few chosen pages with client examples the clients have approved, tested integration details, honest limits and what happens when a person needs to step in. Use clearly labelled demonstrations where no real deployment can be shown. |
| Some pages overlapped. Pairs of service pages answered nearly the same question. | Give each page one distinct job, or merge them after checking which has the links and clicks. We called this potential intent overlap rather than cannibalisation, because we had not shown the pages were taking clicks from each other. |
| Defects on the way to proof. A case studies page linked from several service pages returned 404. An image returned 404. One page logged a rendering error. | Restore or repoint the missing destinations, investigate the rendering error, and check each fix on the live site after it ships. |
| Articles only a click could reach. Blog pagination changed the list without changing the URL, and several articles in the sitemap had no links from any page. | Give paginated pages their own URLs and link each article from relevant pages. These were specific faults, not evidence that the site could not be indexed. |
| The mobile homepage was slow. In two mobile Lighthouse runs the largest content took more than 7 seconds to appear, against a near-perfect desktop score. The slowest element was an animated heading. | Show the heading immediately and animate afterwards, cut script work in the opening screen, and retest several times. These were lab tests, and we did not claim they were costing rankings. |
| Some claims needed support. A few outcome statistics had no source, pricing examples were not full costs, and some security assurances needed qualifying. | Source the results, separate measured outcomes from examples, explain total cost, and publish honest details about working across time zones, communication, ownership and handover. |
| AI search did not need more markup. The site already had structured data, FAQs, videos and an llms.txt file. What it lacked was a baseline. | Make answers and company facts clear and attributable, then track a fixed set of buyer prompts, the pages cited and any referrals, the way we describe here. |
Several of these turn out to be common. A broken link on a page buyers use and a claim the page could not support came up in every audit we pooled for our site audit findings.
What we will measure
Not total impressions. They were already rising, and they were the wrong number. Instead:
- Impressions and clicks from the chosen market, for the chosen searches, on the rewritten pages
- Enquiries from those pages, and whether they were the right kind of buyer
- A fixed set of buyer prompts run through AI tools on a schedule, recording whether the business is cited and which page is used
- The same mobile tests, run several times, before and after the homepage changes
No one can guarantee rankings or AI recommendations, and we have not promised either.
Where it stands
The audit is done and the engagement has started. The first step is choosing one primary market and one or two offers, based on project value, what the team can deliver, existing client proof and how hard the searches are to compete for. When changes have shipped and enough time has passed to measure them fairly, this page will be updated with what happened, whatever it is.
If your site is being seen and not chosen, get in touch.
Client not named
This record describes a current client without naming them, and leaves out their industry, location and figures so it cannot be traced back to them.
Need a practical answer?
See what is limiting your visibility.
Send the URL and tell us what you sell. You will get the first issue worth fixing, with the reason behind it.
Request an audit ↗Free · No call required