The mobile version has less or different content
In the reportPages serve materially less content to the smartphone crawler
mobile_content_parity_gap · Issue
This appears when the main content in smartphone HTML is substantially shorter or different, and the difference repeats beyond normal within-profile variation. Compare the saved content samples and restore any unintended omission in the mobile output.
Severity: High. Review and fix unintended omissions soon: people receiving the mobile response may miss important content.
Why it matters
Search engines recommend equivalent main content on mobile and desktop. The comparison detects a server-response difference; it does not establish what content was indexed or whether the omitted text is important.
What to do
- Compare the saved desktop and smartphone word counts, similarity, variation and missing-text samples. Samples are bounded.
- Find the template, plugin or server rule that omits content for smartphones. Keep the intended main content in the mobile HTML; collapsed content still counts when it remains in that HTML.
- Run another audit to obtain a sufficient comparison of the same page.
What we read
We compare smartphone Chrome with desktop Chrome using two observations of each request profile. The smartphone crawl response supplies the first observation; we then collect a desktop response and a confirmation for each profile. These are browser request profiles, not observations of Googlebot.
The audit freezes a sample before comparison. Its size is the smaller of the eligible page count and the square root of the admitted representative page count, rounded up and limited to 10–100. Selection prioritises an eligible homepage, unresolved rechecks including dismissed findings, then pages from different sections. This does not guarantee every section or template.
A blocked request, cancellation, exhausted comparison budget, incomplete required artifact or challenge leaves the comparison unknown. The content and directive checks require four usable 200 HTML responses. A confirmed smartphone-only status failure belongs to the separate status check and leaves the other three checks unknown.
We compare extracted main content from server HTML without running JavaScript or applying CSS. The word drop must exceed 20% in both rounds, or distinct-token Dice similarity must be 80% or less in both rounds. Cross-profile distance must also exceed the greatest within-profile variation by more than 0.1. Within-profile variation over 5% or missing desktop comparison input leaves the result unknown. With complete, stable comparison inputs, a detected script shell is not applicable.
Example & evidence
Illustrative example — two repeated browser-profile responses for a recipe, not a live audit.
Desktop Chrome receives the whole recipe, 640 words, in both observations:
<main><h1>Sourdough loaf</h1><h2>Ingredients</h2><ul>…</ul><h2>Method</h2><ol>…eight steps…</ol></main>Smartphone Chrome receives 230 words in both observations, with the method replaced by one line:
<h2>Method</h2><p>Open this recipe on a computer to see the method.</p>The word drop exceeds 20%, token overlap is low and neither profile changes between its two observations. The page is reported.
A possible correction: retain the full method in the smartphone HTML, including when its display is collapsed.
Confirm & check again
Check for yourself: Compare the server HTML returned for both browser profiles, not only what the screen shows. Check whether the mobile response contains the same main text; CSS-hidden text still exists in HTML.
Check again: Full audit — Run another audit after correcting the response. The finding closes only when sufficient retained evidence gives a clean result for the same requested page and native final-URL identity. An unread or unselected comparison does not establish a fix. Moving to another final URL, or serving the old final URL through another page, does not fix the original finding. Dismissal is separate from technical state and does not remove the page from recheck consideration.
Limits & sources
The reported sample N of M describes the frozen selection among admitted representative pages in this audit, not every page on the website. Selected pages with incomplete evidence remain in N. An unavailable sample has an explicit reason; it is not a zero-page success.
These requests do not establish what a search crawler received or how it indexed the page. Responses can differ by time, location or other request properties. Missing evidence never counts as a pass.
This is a static HTML heuristic without rendering confirmation. JavaScript changes after loading are not measured; CSS-hidden text still counts. Word counts and token overlap are indicators of a difference, not proof of its importance.
Confidence: Medium because word counts and token similarity are heuristic measures, even when a difference is stable across both rounds.
Sources
- Google, Mobile site and mobile-first indexing best practices — says Google indexes and ranks the mobile version and asks for the same content on mobile and desktop. Checked 25 September 2026.
Editorial review: 3 November 2026