Why pixel width, and not character count
Almost every piece of advice on title length is given in characters, and almost all of it is wrong at the edges. Google renders results in a proportional typeface, so the space a title occupies depends on which letters are in it. A title of 60 narrow lowercase characters can fit comfortably where a title of 50 characters in capitals will not.
The practical consequence is that two titles of identical length can behave completely differently, and that a brand name full of capitals eats considerably more of your budget than its character count suggests. The preview above estimates width using a per-character table rather than counting, which gets much closer to where the cut actually falls.
It remains an estimate. Nothing outside Google can be exact, because the rendering depends on the user’s device, the exact font loaded, and what else Google decides to show for that query. The aim is to be usefully closer than a character count, not to be the renderer.
Writing a title that survives
Put the distinguishing words first. If the end is at risk of truncation, the end should carry the least important information. A brand name at the front costs you the same width on every page you own, which is why most sites put it at the end.
Describe the page, do not list keywords. Google overrides titles it judges unhelpful, and keyword stacking is one of the more reliable triggers. A title that reads like a sentence a person wrote is both more likely to survive and more likely to be clicked.
Make it unique. Duplicate titles across a large page set are a common site-wide problem and they make it harder for a search engine to distinguish those pages. Templates are fine when the variables actually differentiate the result.
Match the page. A title that promises something the page does not deliver earns a click and loses the visitor, which is worse for you than not ranking. The return to the results page is a behaviour you can see in your own analytics if you look for it.
The description is an advert you are not paying for
Meta descriptions have not been a ranking factor for a long time, and a lot of teams stopped writing them carefully as a result. That is a mistake of a different kind: the description frequently decides which of ten adequate results gets the click, and clicks are the entire point of the ranking you worked for.
Google rewrites descriptions often, usually by lifting a passage from the page that matches the query more closely than your written one. This is not a failure. It means the page contains a better answer for that query than the summary did, which is a reasonable outcome. Writing a good description is still worth it for the cases where yours is used.
The most effective descriptions answer the implied question and set an accurate expectation of what the page contains. The least effective repeat the title in slightly different words and add a list of keywords nobody reads.
What changes when an AI answer sits above the result
An assistant answer does not display your title or description at all, so their direct role shrinks on queries that resolve to a generated summary. Their indirect role does not. A clear, specific title is one of the signals that helps any system work out what a page is about, and the discipline of saying what something is in very few words is the same discipline that produces quotable content.
The practical advice is unchanged, with a shift in emphasis. Informational queries are where AI answers appear most and where the click is least reliable, which is an argument for weighting your attention towards the commercial and comparison pages where the click still happens and still converts. Those are the titles worth labouring over.
If you want the fuller version of that argument, we have written it up in SEO vs GEO and in AI SEO vs traditional SEO.