How to judge whether your SEO is actually working

Most businesses paying for SEO share a quiet anxiety: they’re not entirely sure how they’d know if it stopped working, or never started. The reports arrive, the charts point encouragingly upward, and the feeling persists. This article is the plain-English answer: what to measure, in what order, the traps that make good work look bad and bad work look good, and a five-minute self-check you can run without anyone’s permission.

The chain, and why order matters

Search performance is a chain with four links, and they move in sequence.

Positions move first: where you rank for the searches that matter commercially. Note the qualifier, because it does the work: positions for terms your buyers actually use, not positions in general. This is the link the underlying work touches directly, so it’s where honest progress appears earliest.

Impressions follow: how often you are seen. As positions rise into territory people actually look at, the times your business appears in front of a searcher multiply. Impressions are the proof that position movement is happening on searches with real volume behind them, which is why a positions-only report can flatter and an impressions line keeps it honest.

Clicks come next: how often being seen becomes a visit. Clicks concentrate brutally at the top of results, which produces the pattern that confuses more businesses than any other: a site moving from page four to page two is making real, measurable progress that produces almost no clicks yet, because the clicks live on page one. That’s not failure. It’s the mechanism, visibly working.

Enquiries come last: how often a visit becomes business. This is the only link the business actually wanted, and the one everything upstream exists to feed. It’s also the link SEO shares with the website itself: traffic arriving at a page that doesn’t convert is the chain working and the site breaking, which is a different diagnosis with a different fix.

Judge each stage of work by the link it can actually move, and insist your reporting shows the whole chain, not just the flattering end of it.

The new trap: visibility without clicks is no longer failure

Here’s the measurement change of the last two years, and it matters enormously. AI answers now sit on a large share of searches, and when they appear, far fewer people click any result. Pew Research Center’s tracking of nearly seventy thousand real searches put it at 8% clicking with an AI summary present against 15% without, a shift documented across every major study and visible in our own client data: one of our accounts grew impressions 41% and improved five positions while clicks fell 21%.

Ten years ago, impressions up and clicks flat meant something was wrong. Today it often means your content is being read inside the answers, influence arriving without a visit, and the research adds a commercial edge: Seer Interactive found brands cited inside AI answers earn 35% more of the clicks that remain than brands that are not. So add a fifth question to your reporting: when the machines answer your buyers’ questions, are you the business they cite? It’s a checkable fact, the answer names you or it doesn’t, and if your reporting can’t answer it, it’s measuring the last decade.

The traps in both directions

Traps that make good work look bad: judging by clicks in month two, when the movement is still at the front of the chain; measuring against last month instead of last year in a seasonal business, which turns every quiet season into a false alarm; counting all traffic instead of the commercial searches the work actually targets, so that losing worthless visits reads as decline; and the decoupling above, where machine-read visibility gets scored as if it were invisibility.

Traps that make bad work look good: ranking reports full of terms nobody commercial searches, position one for a phrase with no buyers behind it; traffic growth that never touches enquiries, volume without value; “impressions are up” with no position or citation evidence behind it, which can simply mean appearing lower on more irrelevant searches; and any report you can’t understand, because opacity is where weak work hides. A report you can’t explain to your own team isn’t reporting. It’s noise.

What a good report actually contains

Having read a lot of the market’s reporting while researching our own, here’s the plain checklist. The whole chain, positions to enquiries, for the agreed commercial terms. Year-on-year context, not just month-on-month. The AI-citation question answered, engine by engine, for your key buyer questions. What was actually done in the period, in sentences a non-specialist can follow. What hasn’t moved yet, said plainly, with the reason and the plan. And nothing you would need a glossary for. The last two items are the credibility test: honest work is comfortable naming what is still in progress, because the chain makes lag an expected feature rather than a confession.

A report, annotated: our own site, seven weeks after relaunch

The checklist above is easier to use once you’ve seen it used, so here’s a real one. It’s our own website, read seven weeks after we relaunched it on 31 July 2026, with the whole chain shown and the reason written next to each line. The figures are ours and they get published either way.

Positions. The number of buyer searches where we held a first-page position with meaningful exposure went from 10 to 93 between one 28-day window and the next. That’s the line that moves first, so it’s the line we read first. Worth saying what it isn’t. It isn’t “average position”, which on a site that’s just gone from a dozen indexed pages to over a hundred gets worse while Google samples the new ones deep in the results. Ours did. A report that led with average position would’ve called this window a decline.

Impressions. From 11,974 to 59,277. The honest reading is that most of that is footprint, not visibility. Thirty-five pages that had no impressions in the previous window have them in this one, and Google is showing them to see what they are. Impressions confirm the position movement is happening on searches with real volume behind them, which is what they’re for. They don’t on their own prove anyone is choosing us, and the report says so.

Clicks. From 21 to 64, and 34 of those were from the UK. The UK number is the one we report, because the rest is non-buyer geography and a report that counted it would be flattering itself. Clicks lag positions, and at seven weeks most of the new pages are still in the twenties and thirties where clicks don’t yet live. That’s the mechanism doing what it does, and the report names it rather than apologising for it.

Enquiries. Reported separately, from form submissions rather than clicks, at the formal re-measurement on 6 October. Seven weeks is too early to claim anything at the enquiry link. A report that did would be reporting hope.

What was done. A rebuild of the estate around one substantial page per service and per place we serve, with the proof on the page and one next step. Written as sentences, not a task list.

What hasn’t moved yet, and why. Our main local term, “seo aylesbury”, lost more than half its impressions in the window, from 1,003 to 395. The cause was ours. Two of our own pages were competing for it because a redirect pointed at the wrong one. We found it in the data, fixed it the same day, and the next read will show whether the right page has taken over. That line is in the report because a report that hides its own mistake is exactly the noise we warned about above.

Read that back against the checklist. Whole chain, in order. Year-on-year context is the one thing missing, and the report says why, which is that the site is seven weeks old. What was done and what hasn’t moved, in plain words. Nothing you’d need a glossary for. That’s the standard, and it isn’t a high one. It’s just the one most reports don’t meet.

The honest timeline question

How long should it take? The honest answer is that it depends on your starting point, your market and the work’s scope, and a credible agency will set that expectation with you at the start, against a plan, so progress is judged against something you agreed rather than something you hoped. What the credible end of the industry agrees on, and says out loud, is the other half: real search results build over months and compound over years, and a promise of the top of Google in weeks is a promise about a system nobody controls.

The five-minute self-check

You can run this today, without anyone’s permission:

1. Open your own Search Console and look at positions and impressions for your money terms over twelve months, not thirty days.
2. Check the chain’s hand-offs. Where positions have reached page one, are clicks following? Where clicks are arriving, are enquiries?
3. Ask an AI assistant your buyers’ question, phrased the way a buyer would ask it, and see if you are named.
4. Compare against a year ago, not last month, especially if your business breathes with the seasons.
5. Then ask the deciding question: would your current reporting have told you any of this? If not, the problem may not be the work. It may be that nobody’s measuring it in a way that would catch either success or failure.

Get the honest read

If you’d like the reading done for you, we’ll read it honestly: the free audit covers exactly this, where you stand on the whole chain and what we’d fix first, reviewed by a person, with you within a working day.