SERP Features: When Your Ranking Holds but the Clicks Are Gone
A page climbs from fourth to first and traffic falls compared with last month. On today's results pages this is not unusual at all. The reason is rarely bad data. It is that the first position moved: an answer box now sits above it, a video plays beside it, a block of related questions follows it, and the screen space left for links is much smaller than it was a few years ago.
So the explanatory power of "ranking" is declining. It describes where you sit in a list of links, while what users see first is everything that is not a link. The more features a results page carries, the fewer clicks the same ranking produces, and none of that loss appears in position data. Position is still first; the clicks simply do not arrive.
This article follows that thread: which features exist and which queries they appear on, why a stable ranking can still lose clicks, how to prepare for each of them, and finally a candid stance on click-through rate plus a checklist. One boundary up front: there is no official checklist of click-through rate techniques. What follows covers mechanics and testable practice; how it works out on your site can only come from your own comparison testing.
1. The results page is no longer ten blue links
1.1 Which features appear on which queries
Features do not appear at random; each maps to a class of intent. Understanding the mapping tells you whether your page has a real opportunity.
- Featured snippets. These appear where an answer can be stated in a sentence or two and the structure is clear: definition queries, step sequences, comparisons, single-value questions. The block sits above the organic results.
- Sitelinks. These mostly appear on brand and navigational queries, where the user is searching for a specific site or brand. Several in-site entry points are laid out beneath the result.
- People Also Ask. These cover nearly every informational query, presenting follow-up questions that expand and, once expanded, usually generate further questions.
- Image packs. These appear on visual topics: appearance, style, places, charts, physical products. On mobile they consume a particularly large share of the screen.
- Video carousels. These appear on demonstrations, reviews, and explainer queries, sourced from pages already recognised as containing video.
- Knowledge panels and local packs. The first serves entities such as people, organisations, places, and things; the second serves queries with local intent.
1.2 What "first" means once the position is compressed
Ranking is relative; what the user experiences is absolute space. When a block is added above the results, every link is pushed down, and the ones pushed hardest are precisely the previously topmost entries, because what they lose is the first screen.
There is a further subtlety worth separating: one page can appear both inside a snippet and in the link list. It may then look like "ranking first but earning few clicks", because the snippet already delivered the answer and there is no need to click through. That is not a failure but a change in presentation form, and it matters to distinguish "shown but not clicked" from "not shown at all".
2. Stable ranking, falling clicks: zero-click satisfaction
2.1 This is a presentation problem, not a ranking problem
What users want when they search is often not a page but an answer. When the results page supplies the answer directly, the click is skipped. That is what a zero-click result means: the need was met without passing through your page.
Misreading it as a ranking problem triggers a chain of wrong actions: rewriting titles, adding keywords, rewriting content, all in an attempt to "move up a few places" when the position never fell. What actually needs attention is the presentation form for that query: whether you can appear inside it, and whether appearing still earns a click. Those two goals sometimes conflict, and chapter six covers the trade-off.
2.2 Confirming that a feature took the clicks
Three steps, no special tooling.
First, fix a set of URLs in the performance report and compare two windows of equal length. The test is: average position essentially unchanged, impressions essentially unchanged, and only click-through rate declining. All three together largely rule out ranking and demand factors.
Second, check the live results page for those URLs' main queries. Two confounders need care: personalisation and region. Use a private window, sign out, and switch region if necessary, and remember that what you see is a snapshot of this moment; features vary with query phrasing.
Third, check whether the query mix shifted. If exposure for those URLs moved from the head term toward question-style long tail, click-through rate falls even with impressions flat, because question-style answers are often supplied on the results page. Reading query-level data is covered in the Search Console guide.
3. Featured snippets
3.1 Preparing the structure
A snippet is not won by markup; it is selected from content structure. Five things can be prepared.
- Question-style subheadings. Use the phrasing readers search with, not the internal taxonomy name.
- A compact answer immediately after. State the conclusion in two or three sentences in the first paragraph under the heading, rather than building context before answering.
- Definition sentences. Constructions like "X is ..." work well for definition queries: one sentence for the concept, detail after.
- Lists and tables. Use ordered lists for steps and tables for comparisons. These structures are readable on their own and easy to quote as a block.
- Early placement. Put the most quotable passage in the front half of the body instead of burying it in section five.
3.2 Do not sacrifice the body to chase a snippet
One pattern is common: opening an article with a string of unrelated one-line answers meant to cover as many phrasings as possible. It usually fails both ways. Readers get a fragment with no logic, and selection remains unstable because what gets quoted keeps changing.
The more important judgement is that a snippet is not a click. Being quoted can mean more visibility or fewer clicks, since the answer already sits on the results page. If your model depends on users entering the page, sacrificing readability for a snippet trades the long term for the short term.
4. Sitelinks and related questions
4.1 Sitelinks come from structure and internal links
Sitelinks are not obtained through a tag or a structured data field. They reflect the site's own structure: whether the hierarchy is clear, whether navigation is stable, whether important pages receive enough internal links, and whether anchor text is consistent.
Three things can be prepared. Put the most important sections in fixed navigation positions. Link core pages from several related pages rather than once from the homepage. Keep anchor text semantically consistent instead of calling the same page "solutions" today and "product capabilities" tomorrow. The design principles behind that structure are in the internal linking strategy guide.
4.2 People Also Ask: every sub-question needs a quotable answer
People Also Ask is a set of questions that can be expanded further. To appear there, a page needs to cover not only the main question but the questions around it.
The effective format is one question, one answer, in small sections: the subheading is the sub-question, the first paragraph under it gives a short quotable answer, and detail follows. The failing pattern buries sub-questions inside long prose, with the answer scattered across three or four paragraphs; such content is hard to quote and hard to scan.
Where do the sub-questions come from? From real queries. Group the long-tail variants of the main query by phrasing and you will usually list a dozen or more. Covering the five or six most relevant is enough; exhaustive coverage is not required. For mining the variants, see the long-tail keyword research guide.
5. Image packs and video carousels
5.1 Image packs
The opportunity sits on visual topics. Four things can be prepared: give key images dedicated image pages or gallery pages so the image itself becomes an indexable landing point; write meaningful alt text and filenames instead of IMG_2031.jpg; control file weight and loading behaviour so images render on mobile networks; and include images in a sitemap so crawlers have a stable discovery path. For image-side optimisation, see the image optimisation and CDN guide.
One expectation needs correcting: an image pack produces visibility, not clicks, because users view the image on the results page. The right expectation is brand exposure and long-tail discovery, not traffic.
5.2 Video carousels
Getting into a video carousel needs two things at once: the search engine knows a video exists here (discovery), and it considers the video eligible to be shown (eligibility). The first comes from a media sitemap and links on the page; the second from VideoObject structured data plus textual context on the page.
The most common failure is doing only half: a page embeds a video with no structured data at all, or the structured data exists while the page says nothing in text about what the video covers. Rules are in the structured data and rich results guide, and the sitemap side is covered in the media sitemap extensions guide. For how to organise the video content itself, see video content SEO.
6. An honest stance on click-through rate
6.1 There is no official checklist
One thing must be said plainly: there is no official checklist of click-through rate techniques. Official search documentation describes presentation mechanics, structured data specifications, and eligibility requirements. It does not publish entries such as "writing titles this way raises clicks by this much". Any claim to hold an official checklist deserves scepticism.
What can be known is mechanism only. Titles and snippets are generated from page content and may be rewritten. Presentation form is decided by query intent and available content. Feature positions are limited, and not every page has a chance on a given query. These mechanics determine what you can prepare; they do not guarantee an outcome.
6.2 Running your own tests
With no universal answer, comparison testing is the only route. Three workable approaches, ordered by reliability.
Matched groups of similar pages. Take a set of pages with similar structure, adjacent topics, and comparable size, change half, leave half alone, and compare the difference in click-through rate between groups. This comes closest to a controlled experiment and filters out most outside factors.
Before and after on the same page. For one page, take an equal-length window before and after a title change and compare click-through rate on the same query set. Its weakness is that seasonality and external change cannot be excluded, so the windows must be long enough and a control group should be read at the same time.
Read only the target query cluster. When judging an edit, restrict the measurement to the target query set rather than the whole page or the whole site. A title rewrite that works on one query cluster while leaving page-level totals unchanged is entirely normal.
All three need one prerequisite: a trustworthy baseline. For how to establish it, see the SEO performance tracking and measurement guide and the SEO marketing metrics guide.
7. Checklist and boundaries
7.1 Checklist
- The change is confirmed as a presentation shift rather than a ranking or demand shift: position flat, impressions flat, click-through rate down.
- The live results page for the target queries has been checked, with the features present and your own placement recorded.
- Core sections use question-style subheadings, each with a quotable short answer in its first paragraph.
- Step content uses ordered lists and comparison content uses tables, with structural information not flattened into prose.
- Navigation is stable, core pages receive multiple internal links, and anchor text is semantically consistent.
- Key images have dedicated landing points, meaningful filenames and
alttext, and are present in a sitemap. - Video pages carry both a media sitemap entry and
VideoObject, with text on the page describing what the video shows. - Every change has a control and an observation window, measured against the target query set.
7.2 Common misjudgements
- Treating a feature as a ranking reward. A feature is an eligibility for display, judged separately from position.
- Damaging readability to win a snippet. A wall of short sentences at the top loses readers and does not stabilise selection either.
- Reading image pack visibility as traffic. Users view the image in place and do not enter the site; the metrics must be read separately.
- Assuming sitelinks can be obtained through markup. They follow structure and internal links, and no corresponding switch exists.
- Judging a title change by site-wide totals. The target cluster's signal is buried under other pages' swings.
7.3 Priority
If you can do only one thing, do the attribution. Correctly assigning falling clicks to presentation form already prevents every wrong action that follows, and it is the first principle this article shares with the content decay and refresh guide.
If you can do a second, restructure subheadings and short answers. It improves readability, snippet opportunity, and related-question coverage at once, the only move on this list that does three jobs without sacrificing quality. The rest rank by return: image and video preparation is a one-off investment that keeps paying, while title experiments require ongoing controls and patience and should come last.
Source: 16IDC original. Crawling and indexing behaviour, SERP presentation mechanics, and the sitemap protocol follow the official Google Search Central documentation and the sitemaps.org protocol.
Original source: 16IDC ↗