Meta Title Generator
Writes title variations from stated patterns, then scores each one on rendered pixel width, how much of your page’s own vocabulary it carries, where the primary term sits, and whether it duplicates a title you already have.
Generator
Idle
The term the page should rank for. Scored on whether it appears and how early.
Used, not decorative. Distinctive terms are extracted from this and each title is scored on how many it carries.
One per line. A suggestion too close to one of these is flagged, because two pages on one intent is the cannibalisation you are trying to avoid.
Counted in the width, since Google renders it.
Runs locally · nothing sent to a server
How it works
The patterns are templates, and I would rather say so. Five of them: entity-led, benefit-led, comparison-led, question-led and specification-led. There is no model here and no dataset of click-through rates. A browser cannot have one.
What the tool does have is measurement, and that is where the value is:
- Pixel width. Google truncates on rendered width, not character count, so the title is drawn to a canvas in Arial at 20px and measured. Your brand suffix is included, because Google renders it.
- Entity coverage. Distinctive terms are pulled from what you paste about the page, and each title is scored on how many it carries. The extracted terms are shown, so you can see what is being counted.
- Front-loading. Where your primary keyword sits, in pixels from the left. A term that starts past roughly 300px is doing less work than it looks.
- Duplication. Each suggestion is compared against titles you already have. Anything above 60% token overlap is flagged, because two pages on one intent is cannibalisation.
Every score is shown per title with its components, so you can disagree with the ranking rather than take it. Nothing you type leaves your browser.
Frequently asked
No. The suggestions come from five stated patterns filled with your own words. An earlier version of this page claimed the titles were weighted by click-through patterns observed across thousands of pages, which was not true of the code and is not true of any tool that runs entirely in a browser. The scoring is real and the generation is templates.
It is tokenised, stripped of stopwords, and the terms that appear more than once become the page’s vocabulary. Each title is then scored on how many of those terms it contains. The list is printed above the results so you can check it. In the previous version this field was passed to the generator and never read.
Because that is what Google does. Two titles at 58 characters can render 70px apart, so a character limit passes one and fails the other for no reason. The container widths, 600px desktop and 400px mobile, are approximations of Google’s layout rather than published figures.
Often not. Titles get rewritten most commonly when they are too long, keyword-stuffed, or repeated across a site. The first and third are what this tool measures. If yours is rewritten anyway, a brand suffix on every page is the usual cause.