Sev Leo is an SEO expert and IT graduate from Lapland University, specializing in technical SEO, search systems, and performance-driven web architecture.
You’re getting plenty of impressions, but the clicks feel random from week to week. You tweak titles, rewrite meta descriptions, add markup—and still watch click-through rate swing as the results page changes around you.
That’s not just a copy problem; it’s a layout problem on Google search pages. When answers, boxes, and enhancements crowd the screen, some queries become click-poor no matter how polished your listing is. This comparison breaks down the on-page CTR levers you control versus the SERP features you can (and can’t) realistically win, then shows how to validate the impact in Search Console before you commit another sprint.
Clicks Are a SERP Layout Problem
Google search pages (the search results pages, or SERPs, where users see ads, AI answers, SERP features, and organic listings for a query) don’t just reflect demand—they cap how many clicks are even available. CTR (click-through rate: clicks divided by impressions; in Search Console, clicks/impressions are only counted when the user can eventually leave the results page for a non-Google destination) is downstream of that layout.
Zero-click reality
SparkToro’s 2024 write-up of Datos’ analysis puts a hard ceiling on what “better copy” can do: 58.5% of U.S. Google searches resulted in zero clicks, and for every 1,000 searches, only 360 clicks went to the open web (non-Google-owned, non-Google-ad-paying properties). In other words, a large share of queries are structurally click-poor.
That’s why your first decision shouldn’t be “on-page sprint or feature sprint.” It should be classification:
If the SERP is click-available, users still need to click out to complete the task (comparisons, deep research, complex purchases). If it’s click-suppressed, the SERP itself satisfies the intent (direct answers, calculators, entity panels, AI summaries), so even perfect titles and snippets are fighting for leftovers.
Quick search clarifications
The Google Search page is the results screen you land on after you type a query into Google in a browser (google.com) or the Google app.
Your pages show in Google Search when Google has indexed them and decides they match a query.
To find a specific page, search Google using your brand/site name plus a distinctive phrase from the page, or paste the URL and search.
To search within a page you’ve opened, use your browser’s Find function (Ctrl+F on Windows/Linux, Cmd+F on Mac).
On-Page CTR Levers You Control
On-page SEO (changes made on the page itself—content, headings, title/meta, internal links, structured data—intended to improve rankings and/or click-through) can influence what shows on Google search pages, but it can’t dictate it. Google treats your HTML as inputs, then generates the search result presentation it thinks fits the query.
Start with the title link—Google’s term for the clickable title shown in a search result (which may differ from your HTML
). Google Search Central is explicit that title links are generated automatically and can be pulled from multiple places beyond the title tag: on-page headings (like your <h1>), prominent on-page text, anchor text, text in links pointing to the page, and even WebSite structured data. The lever you control is consistency: make the title tag, the main heading, and the page’s “first-glance” text agree on what the page is.</p>
<p>Next is the <strong>snippet</strong>—the descriptive text shown under the title link in an organic result, generated from page content and/or your meta description. Google may choose your <meta name="description"> when it provides a more accurate description than on-page text, which makes meta descriptions an input worth writing (see this <a href="https://skribra.com/seo-guide">SEO guide for on-page basics</a>). But it’s still an input: Google can ignore it and build the snippet from the page instead.</p>
<p>If you need tighter control over what can be shown, Google supports snippet preview controls: <strong>nosnippet / max-snippet / data-nosnippet</strong> (Google-supported controls to limit or block snippet previews, with important exceptions/limits, especially around structured data and feature behavior). max-snippet can limit text preview length, but Google’s robots meta tag documentation notes it doesn’t apply when the information is provided via structured data for rich results. And even perfect markup doesn’t force an enhanced display—Google’s structured data guidelines say structured data can enable a feature, but it’s not guaranteed to show.</p>
<p>The practical line: on-page work can heavily influence the candidate text Google can pull, but the SERP is still an edited presentation, not a page render.</p>
<p>On-page SEO (changes made on the page itself—content, headings, title/meta, internal links, structured data—intended to improve rankings and/or click-through) can influence what shows on Google search pages, but it can’t <em>dictate</em> it. Google treats your HTML as inputs, then generates the search result presentation it thinks fits the query.</p>
<p>Start with the <strong>title link</strong>—Google’s term for the clickable title shown in a search result (which may differ from your HTML <title>). Google Search Central is explicit that <a href="https://developers.google.com/search/docs/appearance/title-link">title links are generated automatically</a> and can be pulled from multiple places beyond the title tag: on-page headings (like your <h1>), prominent on-page text, anchor text, text in links pointing to the page, and even WebSite structured data. The lever you control is consistency: make the title tag, the main heading, and the page’s “first-glance” text agree on what the page is.</p>
<p>Next is the <strong>snippet</strong>—the descriptive text shown under the title link in an organic result, generated from page content and/or your meta description. Google may choose your <meta name="description"> when it provides a more accurate description than on-page text, which makes meta descriptions an input worth writing (see this <a href="https://skribra.com/seo-guide">SEO guide for on-page basics</a>). But it’s still an input: Google can ignore it and build the snippet from the page instead.</p>
<p>If you need tighter control over what can be shown, Google supports snippet preview controls: <strong>nosnippet / max-snippet / data-nosnippet</strong> (Google-supported controls to limit or block snippet previews, with important exceptions/limits, especially around structured data and feature behavior). max-snippet can limit text preview length, but Google’s robots meta tag documentation notes it doesn’t apply when the information is provided via structured data for rich results. And even perfect markup doesn’t force an enhanced display—Google’s structured data guidelines say structured data can enable a feature, but it’s not guaranteed to show.</p>
<p>The practical line: on-page work can heavily influence the candidate text Google can pull, but the SERP is still an edited presentation, not a page render.</p>
<p><img src="https://assets.skribra.com/images/6954b95483462c019ddf8ff7/2026-09-12T07:56:45.468Z.jpg" alt="SEO workstation with SERP mockup; magenta label on the screen reads "title link" to highlight CTR control."></p>
<h2>SERP Features: Winnable vs Unwinnable</h2>
<p>On Google search pages, <strong>SERP features</strong> (non-standard result elements like featured snippets, sitelinks, knowledge panels, local packs, and AI Overviews that change what appears on the results page) split into two ownership buckets: what you can sometimes earn, and what you mostly have to live with.</p>
<p><strong>Winnable (you can sometimes own the real estate):</strong><br>
A <strong>featured snippet</strong> (a special result box that shows the snippet first—often called “position zero”; Google chooses whether to elevate a page) is “winnable,” but not markable: Google is explicit that you can’t tag a page into being the featured snippet, and the only guaranteed way to prevent your content from showing as one is to use nosnippet. Sitelinks are similar: <strong>sitelinks</strong> are automated, and Google only shows them when it thinks they’ll be useful.</p>
<p>A <strong>rich result</strong> (an enhanced organic result that uses structured data, like job postings or events, and is reported via Search appearance/rich-result reports when applicable) is also in the “sometimes” category—but for two concrete reasons. First, structured data is eligibility, not a promise: Google’s structured data guidelines say it doesn’t guarantee your structured data will show in Search results. Second, eligibility can disappear: in June 2025, Google announced multiple structured-data features had been phased out from Search results (Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement, Vehicle Listing), with removal from Search Console rich result reporting and related tooling starting September 9.</p>
<p><strong>Mostly unwinnable (Google-owned, or opt-out):</strong><br>
<strong>AI Overviews / AI Mode</strong> (Google Search generative AI features that synthesize answers on the SERP and may include citations/links to sources) aren’t something you “win” the way you win an organic listing. Google provides a <strong>Search generative AI control</strong> in Search Console that lets you exclude your site from AI Overviews, AI Mode, and related features; if you exclude, you won’t receive impressions or traffic from those features.</p>
<p>That ownership test is the shortcut: for each query, ask whether the feature is one you can realistically earn—or one you must decide to tolerate or opt out of.</p>
<p>On Google search pages, <strong><a href="https://skribra.com/glossary/serp">SERP features</a></strong> (non-standard result elements like featured snippets, sitelinks, knowledge panels, local packs, and AI Overviews that change what appears on the results page) split into two ownership buckets: what you can sometimes earn, and what you mostly have to live with.</p>
<p><strong>Winnable (you can sometimes own the real estate):</strong><br>
A <strong>featured snippet</strong> (a special result box that shows the snippet first—often called “position zero”; Google chooses whether to elevate a page) is “winnable,” but not markable: Google is explicit that you can’t tag a page into being the featured snippet, and the only guaranteed way to prevent your content from showing as one is to use nosnippet. Sitelinks are similar: <strong>sitelinks</strong> are automated, and Google only shows them when it thinks they’ll be useful.</p>
<p>A <strong>rich result</strong> (an enhanced organic result that uses structured data, like job postings or events, and is reported via Search appearance/rich-result reports when applicable) is also in the “sometimes” category—but for two concrete reasons. First, structured data is eligibility, not a promise: Google’s structured data guidelines say it doesn’t guarantee your structured data will show in Search results. Second, eligibility can disappear: in June 2025, Google announced multiple structured-data features had been phased out from Search results (Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement, Vehicle Listing), with <a href="https://developers.google.com/search/blog/2025/06/simplifying-search-results">removal from Search Console reporting</a> and related tooling starting September 9.</p>
<p><strong>Mostly unwinnable (Google-owned, or opt-out):</strong><br>
<strong>AI Overviews / AI Mode</strong> (Google Search generative AI features that synthesize answers on the SERP and may include citations/links to sources) aren’t something you “win” the way you win an organic listing. Google provides a <strong>Search generative AI control</strong> in Search Console that lets you exclude your site from AI Overviews, AI Mode, and related features; if you exclude, you won’t receive impressions or traffic from those features.</p>
<p>That ownership test is the shortcut: for each query, ask whether the feature is one you can realistically earn—or one you must decide to tolerate or opt out of.</p>
<h2>Head-to-Head: Which Wins Clicks</h2>
<p>On google search pages, “on-page vs SERP features” isn’t a vibe choice. It’s a per-query bet: are you trying to earn a layout that historically pulls more clicks, or tightening what you control and avoiding volatile real estate?</p>
<table>
<thead>
<tr>
<th>Criterion</th>
<th>On-page SEO</th>
<th>SERP features</th>
<th>Winner (for clicks)</th>
</tr>
</thead>
<tbody>
<tr>
<td>CTR evidence at Position 1</td>
<td>“Purely organic” P1: 34.2%</td>
<td>P1 with sitelinks: 46.9%; P1 with featured snippet: 23.3%</td>
<td>SERP features (sitelinks); on-page over snippets</td>
</tr>
<tr>
<td>AI answers and external clicks</td>
<td>Can’t force removal</td>
<td>AI Overview citation clicks: ~1% of visits; removing AI Overviews: +8.8pp external CTR</td>
<td>SERP features (opt out / avoid)</td>
</tr>
<tr>
<td>Controllability</td>
<td>High: you ship the page</td>
<td>Medium: Google decides display</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>Eligibility / deprecation risk</td>
<td>Lower: page remains eligible</td>
<td>Structured-data types phased out (2025-06-12)</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>“Guarantee” risk</td>
<td>Inputs, not guarantees</td>
<td>Structured data display not guaranteed; featured snippet not markable</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>Measurement cleanliness</td>
<td>URL-level testing possible</td>
<td>Search appearance attribution not guaranteed per URL</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>Scenario: you can win sitelinks at P1</td>
<td>Strengthen IA and targeting</td>
<td>Highest observed P1 CTR (46.9%)</td>
<td>SERP feature sprint</td>
</tr>
<tr>
<td>Scenario: SERP shows featured snippet / AI answer</td>
<td>Improve the organic listing</td>
<td>Use nosnippet / AI control; AI control changes: 1–2 days</td>
<td>Depends: opt-out beats chase</td>
</tr>
</tbody>
</table>
<p>If the feature demonstrably lifts CTR (sitelinks), chase it; if it compresses clicks (featured snippets, AI answers), spend on controllable listings—or opt out fast and measure.</p>
<iframe width="560" height="315" src="https://www.youtube.com/embed/1yhPX7Q7KOQ" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
<p>On google search pages, “on-page vs SERP features” isn’t a vibe choice. It’s a per-query bet: are you trying to earn a layout that historically pulls more clicks, or tightening what you control and avoiding volatile real estate?</p>
<table>
<thead>
<tr>
<th>Criterion</th>
<th>On-page SEO</th>
<th>SERP features</th>
<th>Winner (for clicks)</th>
</tr>
</thead>
<tbody>
<tr>
<td>CTR evidence at Position 1</td>
<td>“Purely organic” P1: 34.2%</td>
<td>P1 with sitelinks: 46.9%; P1 with featured snippet: 23.3%</td>
<td>SERP features (sitelinks); on-page over snippets</td>
</tr>
<tr>
<td>AI answers and external clicks</td>
<td>Can’t force removal</td>
<td>AI Overview citation clicks: ~1% of visits; removing AI Overviews: +8.8pp external CTR</td>
<td>SERP features (opt out / avoid)</td>
</tr>
<tr>
<td>Controllability</td>
<td>High: you ship the page</td>
<td>Medium: Google decides display</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>Eligibility / deprecation risk</td>
<td>Lower: page remains eligible</td>
<td>Structured-data types phased out (2025-06-12)</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>“Guarantee” risk</td>
<td>Inputs, not guarantees</td>
<td>Structured data display not guaranteed; featured snippet not markable</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>Measurement cleanliness</td>
<td>URL-level testing possible</td>
<td>Search appearance attribution not guaranteed per URL</td>
<td>On-page SEO</td>
</tr>
<tr>
<td>Scenario: you can win sitelinks at P1</td>
<td>Strengthen IA and targeting</td>
<td>Highest observed P1 CTR (46.9%)</td>
<td>SERP feature sprint</td>
</tr>
<tr>
<td>Scenario: SERP shows featured snippet / AI answer</td>
<td>Improve the organic listing</td>
<td>Use nosnippet / AI control; AI control changes: 1–2 days</td>
<td>Depends: opt-out beats chase</td>
</tr>
</tbody>
</table>
<p>If the feature demonstrably lifts CTR (sitelinks), chase it; if it compresses clicks (featured snippets, AI answers), spend on controllable listings—or opt out fast and measure.</p>
<p><img src="https://assets.skribra.com/images/6954b95483462c019ddf8ff7/2026-09-12T07:57:20.924Z.jpg" alt="Comparison: On-page SEO vs SERP features across Sitelinks and Featured snippet outcomes for clicks"></p>
<h2>Prove It in Search Console</h2>
<p>You’re trying to answer one question: did clicks move because you changed the page, or because Google changed the layout on google search pages?</p>
<ol>
<li>
<p><strong>Lock the test set (so you’re not measuring a moving target).</strong> Pick a small set of URLs you will change, and a matched set you will not change. In Search Console’s <strong>Performance report (Search results)</strong>, filter to those pages and save the view so you can reuse the exact filters later.</p>
</li>
<li>
<p><strong>Segment by Search appearance before you look at CTR.</strong> <strong>Search appearance</strong> (a Performance-report dimension/filter that groups clicks and impressions by certain result types/features, with attribution limitations) is where you separate “plain blue link” behavior from feature-influenced behavior. Create separate views for the appearances you care about (for example: rich-result appearances vs everything else), and keep them as distinct comparisons.</p>
</li>
<li>
<p><strong>Treat Search appearance as directional, not forensic.</strong> Search Console explicitly warns that if you filter by Search appearance, it’s <strong>not guaranteed</strong> that all clicks for a given URL came from the filtered feature type, because clicks are assigned to a URL rather than a specific (URL + feature) combination. Use the dimension to spot shifts, then sanity-check by drilling into queries/pages instead of declaring feature-level causality from one filtered chart.</p>
</li>
<li>
<p><strong>Measure generative AI separately, using the dedicated report.</strong> Use the <strong>Generative AI performance report (Search)</strong> to isolate how your site performs in generative AI features like AI Overviews / AI Mode, then compare the same before/after windows you’re using in the Search results Performance report.</p>
</li>
<li>
<p><strong>Run clean before/after windows, then decide the next sprint.</strong> If you toggled the <strong>Search generative AI control</strong>, treat the toggle date as the intervention point (excluded sites receive no impressions or traffic from those features), and keep your before/after windows anchored to it. If CTR only moves inside a specific appearance (or only inside the generative AI report), you’re looking at a feature shift; if it moves similarly across appearances, it’s a better case for an on-page sprint.</p>
</li>
</ol>
<p>You’re trying to answer one question: did clicks move because you changed the page, or because Google changed the layout on google search pages?</p>
<ol>
<li>
<p><strong>Lock the test set (so you’re not measuring a moving target).</strong> Pick a small set of URLs you will change, and a matched set you will not change. In Search Console’s <strong>Performance report (Search results)</strong>, filter to those pages and save the view so you can reuse the exact filters later.</p>
</li>
<li>
<p><strong>Segment by Search appearance before you look at CTR.</strong> <strong>Search appearance</strong> (a Performance-report dimension/filter that groups clicks and impressions by certain result types/features, with attribution limitations) is where you separate “plain blue link” behavior from feature-influenced behavior. Create separate views for the appearances you care about (for example: rich-result appearances vs everything else), and keep them as distinct comparisons.</p>
</li>
<li>
<p><strong>Treat Search appearance as directional, not forensic.</strong> Search Console explicitly warns that if you filter by Search appearance, it’s <strong>not guaranteed</strong> that all clicks for a given URL came from the filtered feature type, because clicks are assigned to a URL rather than a specific (URL + feature) combination. Use the dimension to spot shifts, then sanity-check by drilling into queries/pages instead of declaring feature-level causality from one filtered chart. (See Search Console’s caveats on the <a href="https://support.google.com/webmasters/answer/17011259">Search appearance dimension</a>.)</p>
</li>
<li>
<p><strong>Measure generative AI separately, using the dedicated report.</strong> Use the <strong>Generative AI performance report (Search)</strong> to isolate how your site performs in generative AI features like AI Overviews / AI Mode, then compare the same before/after windows you’re using in the Search results Performance report.</p>
</li>
<li>
<p><strong>Run clean before/after windows, then decide the next sprint.</strong> If you toggled the <strong>Search generative AI control</strong>, treat the toggle date as the intervention point (excluded sites receive no impressions or traffic from those features), and keep your before/after windows anchored to it. If CTR only moves inside a specific appearance (or only inside the generative AI report), you’re looking at a feature shift; if it moves similarly across appearances, it’s a better case for an on-page sprint.</p>
</li>
</ol>
<h2>Pick the sprint per SERP</h2>
<p>If your CTR is swinging while impressions hold, assume the Google search pages changed first—and treat “better copy” as a second-order lever. Classify the queries you care about as click-available or click-suppressed, then place a deliberate bet: pursue features that expand clickable real estate (like sitelinks), and stop chasing features that compress clicks—use controls like nosnippet or the Search generative AI control when the layout is the problem. Your next step is to lock a small test set in Search Console, segment by Search appearance (and check the Generative AI performance report separately), then choose the next sprint based on where the movement actually shows up.</p>
<section class="faq">
<h2>Frequently Asked Questions</h2>
<dl>
<dt>Why does Google rewrite my title link and snippet on Google search pages even when my title tag and meta description look perfect?</dt>
<dd>Because Google generates the title link and snippet dynamically and may pull them from on-page headings, prominent text, anchor text, or other signals—not just your HTML title tag and meta description. Treat your HTML as inputs and validate what actually shows on the SERP before judging CTR changes.</dd>
<dt>How do I stop my site from showing in AI Overviews on Google search pages, and what happens if I opt out?</dt>
<dd>Use the Search generative AI control in Google Search Console to exclude your site from AI Overviews, AI Mode, and related features. If you exclude, your site receives no impressions or traffic from those generative AI features.</dd>
<dt>If AI Overviews are showing for my queries, what click impact should I expect on Google search pages?</dt>
<dd>A preregistered randomized field experiment on Google Search reported that removing AI Overviews increased external click-through by 8.8 percentage points, and that AI Mode-only reduced click-through rate versus current Search. Use this as a testable hypothesis for your own before/after measurement.</dd>
<dt>Do people actually click the citation links inside AI Overviews on Google search pages?</dt>
<dd>A 2026 arXiv analysis of web-browsing data reported that clicks to sources cited in AI Overviews were very rare—occurring in only about 1% of visits to AI Overviews. Measure your own generative-AI referrals separately rather than assuming citations drive meaningful traffic.</dd>
<dt>Are People Also Ask boxes taking clicks away from my organic results on Google search pages?</dt>
<dd>People Also Ask expands the SERP and can satisfy intent without a click by letting users keep exploring questions inside Google search pages. The actionable step is to segment performance by query and watch for CTR drops when People Also Ask appears for the same terms.</dd>
</dl>
</section>
<section class="cta">
<hr />
<h2>Turn SERP bets into output</h2>
<p>Once you’ve picked the next sprint per query, the bottleneck is shipping the right pages fast enough to test and iterate before the layout shifts again.</p>
<p><a href="https://skribra.com/">Skribra</a> is an AI-driven SEO content system that builds a rolling 30-day content calendar, writes and publishes long-form articles, and updates live posts based on performance data—so your next CTR experiments actually get executed. Start with the 3-Day Free Trial.</p>
</section>
Written by
Skribra
This article was crafted with AI-powered content generation. Skribra creates SEO-optimized articles that rank.
