The short version
- Position alone is no longer a sufficient measure: what matters is how much space is left above you.
- Two queries at position one can be worth ten times each other, depending on which blocks are present.
- Expected CTR should be calculated per SERP type, not per average position.
- Generated answers shift the competition from position to citation.
- Anyone not auditing the blocks on their own queries is measuring half the outcome.
For fifteen years the results page was a list. Ten blue links, a few ads on top, a clear order. Optimising meant climbing that list, and position was a decent proxy for traffic.
Today the same query can return a local pack, a generated answer box, an image carousel, a related-questions section, three ads and finally — below the fold — the first organic result. Position one still exists, but it no longer means the same thing.
This article is about a discipline almost nobody practises: auditing what your SERPs actually look like, understanding how much space genuinely remains, and deciding where to invest accordingly. It is not theory — it changes which queries are worth chasing and which are not.
Why average position misleads
The number in every report — average position — hides two things at once.
The first is dispersion: an average of 8 can be position 3 for 40% of impressions and 15 for the rest, depending on where the searcher is. On local queries this variance is enormous, and in a city like Naples, where the commercial fabric changes radically between the historic centre, the hill districts and the western suburbs, the average often describes a situation that no real user experiences.
The second, less discussed, is that position tells you where you are in the organic list, not where you are on the page. With four ads, a local pack and a generated box above you, organic position one can sit two screens from the top.
Positions identical to three months ago, impressions steady, clicks down 30%. You did not lose ranking: you lost space. The only way to notice is to look at SERP composition, because in standard reports this scenario is indistinguishable from a content problem.
Auditing the blocks: a two-hour exercise
The exercise is simple and almost always reorders priorities. Take your tracked query set and, for each one, note which blocks appear.
| Block | Effect on organic clicks | Can you compete? |
|---|---|---|
| Local pack | Very strong: absorbs most commercial clicks | Yes, via the business listing rather than the website |
| Generated answer box | Strong and growing: reduces clicks for everyone | Indirectly, with citable content |
| Featured snippet | Concentrates clicks on a single result | Yes, with structure and a short answer |
| Related questions | Moderate: lengthens the page, distracts | Yes, by answering explicitly |
| Ads | Proportional to count, strong on commercial queries | Only by paying |
| Carousels and images | Variable, strong on visual queries | Yes, with your own optimised images |
The result is nearly always the same: your query set splits into three very different groups that until now were measured together. Queries where organic results still dominate and content investment pays. Queries where the local pack rules and the right investment is the listing and reviews. And queries where everything sits above you and organic return is structurally low, whatever you do.
Expected CTR by SERP type, not by position
The correct way to judge performance is not against a generic position table but against what is reasonable for that kind of SERP.
These are orders of magnitude rather than constants, varying by sector and intent. But the point is structural — a 6% click-through at position one is not a copywriting failure when a generated box sits above you. It is that SERP's normal value.
The practical consequence concerns decisions. Before rewriting the title and description of an underperforming page, check whether the problem is your result or the space it lives in. In the first case a rewrite delivers 10% to 25% more clicks; in the second it delivers nothing and the time belonged elsewhere.
Generated answers: from position to citation
The deepest change of the past two years is that a growing share of searches never ends in a click. The user receives a paragraph with two or three cited sources and acts on it.
That moves the competition to a different plane. It is no longer about being first in a list but about being the source the system uses to build the answer. The mechanics differ: there is no position, there is a citation frequency; there is no keyword, there is a conversational question loaded with conditions.
What makes content citable, in practice: answering the question in the first paragraph without preamble; containing verifiable data — prices, timeframes, conditions, numbers; stating constraints explicitly, meaning areas served, hours, requirements; and being structured in short blocks with headings that mirror the real question. Nothing esoteric: it is the same writing that works for a hurried human reader.
The new part is measurement. Until recently there was no way to know whether or how often you were cited. The AI Analytics module measures mentions, citations and sentiment across the questions in your category, and lets you compare against competitors — which is the only figure that makes your own number interpretable.
In local categories we still routinely find clearly commercial questions where no business from the city appears convincingly: answers fall back on generic criteria or national portals. Whoever publishes specific, verifiable material first occupies that space for a modest investment. It will not last.
When chasing a featured snippet is worth it
The featured snippet is the only block a small site can take from a large competitor in a few hours of work, which is why it deserves a method.
Three conditions have to hold together: the query has an active snippet; it is held by someone you match or beat organically; and the question has a short, factual answer rather than an opinion.
The work is reformulating the relevant block of your page: a heading mirroring the exact question, followed immediately by a forty-to-sixty-word answer that is self-contained, with no dependency on the following paragraph. If the answer is a procedure, a numbered list; if it is a comparison, a table.
The flip side deserves saying too: winning a snippet does not always increase clicks. On short factual questions the user reads and stops. It is worth chasing when the brief answer does not resolve the need — that is, when the reader still has to come to you in order to act.
The title Google rewrites
One last element affects clicks and almost nobody checks it: the title shown in the results is not always the one you wrote. Google rewrites it when it judges yours a poor description of the page, and it does so more often than people assume.
The most frequent rewrites follow three patterns. The title is too long and gets truncated or replaced with a fragment of the content. The title is stuffed with repetitions of the same keyword and gets replaced by the H1. Or the title is generic — "Services", "Home" — and gets replaced with the business name plus something lifted from the page.
The practical problem is that a rewrite takes control away from you on precisely the element that decides the click. There is no way to prevent it outright, but it becomes far rarer when three conditions hold: the title contains the real phrasing of the question, it fits within the 55-60 usable characters on mobile, and it honestly describes what is on the page.
Checking is trivial and nobody does it: search for ten of your main pages and compare the displayed title with the one you wrote. Where they differ, you have found pages where you are letting the engine decide. For a local business, where the gap between a generic title and one naming the district and the service can be worth several points of click-through, it is one of the best time-to-result checks available.
How this changes query selection
- Split queries by SERP typeThree distinct tracking groups: organic-dominant, pack-dominant, saturated. Measuring them together produces averages that describe nothing.
- Move budget towards the first groupThat is where content produces return proportional to effort. These are usually longer, more specific queries with less volume and more intent.
- On the second group, work the listingPrimary category, completeness, recent reviews, real photographs. Operational work that pays in weeks rather than months.
- On the third group, decide deliberatelyEither accept a low organic return, cover it with paid, or drop it. Continuing to invest content while hoping the space returns is the worst of the three.
- Redo the audit quarterlySERP composition changes. A query that is clean today may carry a generated block in six months, and your plan should notice before the revenue does.
The monitoring this requires
Doing all of the above needs three data points read together, which in most setups live in separate tools with different periods.
Position by query group, not aggregated, to see real movement. Clicks and impressions per URL, to catch the divergence between stable positions and falling clicks, which is the signature of a SERP change. Presence in generated answers, for the layer the other two cannot see.
Holding them in one place is not a convenience question: it is the condition for being able to say with confidence "positions are steady, clicks fell, therefore the results page changed". With three separate tools and three definitions of "last 30 days", that sentence takes a week of work instead of a glance. In the free layer of Semalt the three share a period and a domain, which is exactly why we use it as the measurement base on projects.
Audit the blocks on your own queries
Positions by group, clicks per URL and AI answer visibility, on one period and one domain. Free, with no time limit.
Sign in to Semalt See Google SERPFrequently asked questions
Does position still matter?
It does, but it has to be read alongside SERP composition. Position 1 on a saturated page is worth less than position 3 on a clean one.
How do I know the SERP changed?
Stable positions with falling clicks on the same impressions. That is the signature: your result did not get worse, the space above it grew.
Do generated boxes really reduce clicks?
On informational content, markedly. On commercial queries with buying intent, far less, because the user still needs to choose a supplier.
Is chasing a featured snippet worthwhile?
When it is held by someone you already outrank organically and the short answer does not resolve the reader's need. Otherwise you may give away the answer without receiving the visit.
How often should the block audit be repeated?
Quarterly. SERP composition changes faster than positions do.
Conclusion
Measuring position alone today means looking at half the picture. The space available above your result has become an independent variable, one that changes without you doing anything and that determines what each position you win is actually worth.
The exercise for this week costs two hours: take your top thirty queries, note which blocks appear, and split the list into the three groups. It almost always produces a budget reallocation that no amount of keyword research would have suggested.
For the ongoing measurement, connect your domain: positions, clicks and generated-answer visibility read together are what make the divergence visible before it becomes a problem. And if you would rather we ran the audit on your queries, get in touch — it is part of the free initial audit we deliver within 24 hours.
