On-page SEO: the part Google can’t rewrite for you

A study most title tag advice ignores

John McAlpin tracked roughly 30,000 keywords through Advanced Web Ranking and crawled the resulting result pages with Screaming Frog, comparing the title tag published in each page's HTML against the title Google actually displayed in search. He published the results in Search Engine Land on 1 May 2025: Google changed the title in 76% of cases. Of the titles it changed, only 35% of the original words survived into the version shown in search.

The pattern wasn't random. Brand names were the single most common casualty, stripped out of 63% of the titles that got modified. Intent mattered too: on informational pages, only 5.35% of modified titles kept the original target keyword at all, against 31.31% on commercial pages. If you're writing keyword-led titles expecting them to reach the search result unchanged, this study says otherwise for roughly three out of four pages.

Meta descriptions, rewritten almost as often

The comparable data on meta descriptions is older and worth flagging as such. Portent analysed 30,000 keywords and published the numbers on 10 September 2020: Google rewrote 71% of meta descriptions on mobile and 68% on desktop, using the publisher's own description roughly three times in ten. Rewrite rates rose for longer, lower-volume queries and for results further down the first page. That data is five years old and nobody appears to have repeated it at this scale since, so treat the exact percentage as dated even if the underlying pattern, that Google treats your meta description as a candidate rather than a command, still holds.

Why Google intervenes, in its own words

Google's title-link documentation, last updated 10 December 2025, describes the process as fully automated, built from the title element, the page's main visible heading, heading tags, the og:title meta tag, prominently styled text, the page's own content and inbound anchor text, and any WebSite structured data. It also names, specifically, when Google steps in: a title that is half-empty or vague, a date in the title that no longer matches the visible content, a title that misrepresents what the page is about, boilerplate text repeated across many pages of the same site, several headings of equal visual prominence competing for attention, or a mismatch between the title's language and the page's.

Google's snippet documentation, last updated 20 April 2026, says the same thing about the description: snippets are primarily generated from the page content itself, and Google uses the meta description tag only when it thinks it will give users a more accurate result than what it can pull from the page. Neither document treats your tag as an instruction. Both treat it as one input the system checks against the page and is free to override.

The on-page work that survives an override

Read the trigger list above again, because it is a checklist, not a warning. Half-empty titles, stale dates, boilerplate repeated across a template, headings of competing prominence: every one of those is fixable, and fixing them is exactly the work our own On-Page Optimization spec does page by page. Writing a specific, accurate title under sixty characters and a description that states one real benefit does not guarantee the string you wrote ships untouched. It does remove every documented reason Google's system has to intervene, which is the most any on-page rewrite can honestly promise.

The other half of that spec is deciding which pages are worth the work at all. Pages sitting at position 8 to 20 for a keyword that actually converts are where a rewrite has somewhere to go; a page with no defined target keyword is not being optimized, it is being guessed at, which is the same conclusion our guide to keyword research reaches from a different direction.

The parts nothing automated touches

Three elements of on-page work sit completely outside the systems described above, and none of them appears in either Google document as something the algorithm generates on your behalf.

Internal links. Nothing in Google's title or snippet pipeline manufactures a link into your page from somewhere else on your own site. A link either exists, with anchor text you chose, or it does not. That makes internal linking one of the few on-page levers with no override risk at all: what you build is what stays built.

Schema markup. Structured data is binary in a way a title never is. It is either present in the page's HTML, correctly typed, or it is absent, and Google does not write JSON-LD for you when you skip it. Our guide to schema that earns citations covers which types are worth the implementation time and which are decoration.

Content that matches what already ranks. A title rewrite cannot manufacture a section your page is missing. If the five results above yours all answer a sub-question yours skips, that gap is what a SERP-level content review exists to find, which is the starting point of our guide to why a competitor keeps outranking you.

A ten-minute check on your own pages

1. Compare the tag against the SERP. Search for one of your own pages in an incognito window, then open view-source on that same page and read the raw <title> element. If the two do not match, Google has already rewritten this specific page, and you now know which pattern from the trigger list to look for.
2. Check for template-wide boilerplate. Open two or three pages built from the same template, product pages or location pages are the usual offenders, and see whether the titles share large blocks of repeated wording beyond the brand name. Google names this explicitly as a rewrite trigger.
3. Count your headings. View-source again and confirm the page has one heading that is clearly the largest and most prominent, not several competing ones. Multiple equally-weighted headings are on Google's own list.
4. Count real internal links into the page. Search your own site for links pointing at the URL with relevant anchor text, not navigation or footer links. Anything under two or three is a gap no algorithm will close on your behalf.

What we could not verify

Neither Google document ranks its listed title-link sources against each other, so a page can satisfy every stated criterion and still be rewritten for a reason Google has not published. The McAlpin study covers one publication's own tracked keyword set rather than a Google-disclosed figure, and the Portent meta description numbers are five years old and have not been publicly repeated at that scale since. Treat the direction of both findings, that your tag is a candidate rather than a command, as solid, and treat the exact percentages as a snapshot rather than a constant.

Sources

The title tag rewrite rate, word retention and brand removal figures come from John McAlpin, Google changed 76% of title tags in Q1 2025, Search Engine Land, 1 May 2025. The rewrite triggers and title-link sourcing come from Google's own Influencing your title links in search results documentation, last updated 10 December 2025. The meta description rewrite rates are from Evan Hall and Matthew Henry, How Often Does Google Rewrite Meta Descriptions?, Portent, 10 September 2020. The snippet generation policy is from Google's Control your snippets in search results documentation, last updated 20 April 2026.

Common questions