Meta Description Generator
Paste your page text and get a description drafted from it, fitted to the width Google renders rather than to a character count.
By opening this website or using its tools, you agree to our Terms and Privacy Policy.
Picks the sentences that best represent the page and fits them to the width Google renders. No model involved — this works in every browser.
Write it instead of extracting it
Optional. Uses your browser's own on-device model — the draft button above needs no model at all.
Checking what this browser can run…
Your file never leaves your browser during processing
How it works
To generate a meta description, paste your page’s text or HTML above and press draft — the sentences that best represent the page are selected and packed to fit roughly 920 pixels, which is where Google truncates a description on desktop. Everything runs in your browser, and the result is editable with a live pixel measurement.
The draft is extractive: sentences from your page are scored on how many of the page’s own recurring content words they contain, with a bonus for appearing early and for containing the keyword you named, and the best ones are packed into the available width and put back in reading order. That is a real draft from the real page rather than a template with your keyword dropped into it, and because it is a plain algorithm it works in every browser with no model and no download.
The fit is measured, not counted. Google truncates a description by rendered pixel width, not character count — a line of narrow letters fits where the same number of wide ones does not — so the width here is measured with the Canvas text-measurement API at roughly Google’s desktop render, the same measurement /text/meta-length uses. The commonly quoted 155-character figure is a rough consequence of that, which is why it is shown as a guide next to the pixel bar rather than as the limit.
Where your browser provides an on-device language model, the second panel rewrites rather than extracts, which is better copy. It goes through exactly the same measurement: a model cannot see rendered pixels and miscounts characters, so anything it produces is measured and trimmed at a word boundary here rather than trusted to fit. Nothing is sent to a server either way — a browser without a model gets the extractive draft and a note saying so, not an error.
Pasting raw HTML is fine. Scripts, styles, navigation, headers, and footers are removed before anything is read, and the markup is parsed in a detached document that never touches this page — the same rule the SVG and Markdown tools follow.
FAQ
How long should a meta description be?
About 145–155 characters is the practical target, but the real limit is width: Google truncates at roughly 920 pixels on desktop and around 120 characters on mobile. The bar above measures the actual rendered width of what you have written, which is the more reliable of the two signals.
Does this need an AI model?
No. The draft button extracts and fits sentences from your own page text using a plain algorithm that runs in every browser. The generative panel is an optional layer that appears where your browser provides an on-device model — Chrome and Edge on desktop currently do; Safari and Firefox do not.
Is my page content uploaded?
No. The extraction, the measurement, and — where a model exists — the generation all run in your browser. The live network counter below the tool shows zero bytes sent during processing.
Will Google use the description I write?
Not always. Google frequently rewrites descriptions using text from the page when it judges another passage a better answer to the query. A good description raises the odds and controls the wording when it is used; it is not a guarantee.
Why does it warn that my description is missing the keyword?
Because Google bolds query terms it finds in the description, and that bolding is much of what makes a result worth reading in a list. A description that never mentions what the page is about usually means the draft drifted off topic.