What FID measured
First Input Delay timed one narrow thing: the gap between a visitor’s first click, tap or key press and the moment the browser was free to start running the code attached to it. If the main thread was busy finishing a script, that wait was the delay. Once the handler began, the measurement stopped.
That stopping point is the whole story of why it was replaced. FID never counted how long your code then took, and never counted the time before anything appeared on screen. A page could start work promptly and leave the visitor staring at an unchanged button for a long time afterwards, and still score well. It also only ever recorded the first interaction, and only if a real person interacted at all, so pages people simply read produced no measurement.
It was retired as a Core Web Vital in March 2024 and replaced by Interaction to Next Paint, which measures interactions throughout a visit and follows each one all the way to the moment the screen visibly updates.
Why FID still matters
It matters now mostly as a warning sign. If an SEO audit, a developer’s report or an agency proposal still quotes First Input Delay as a current metric, the document was either written from an out-of-date template or generated by a tool that has not been updated. That tells you something about how carefully the rest of it was assembled, and every recommendation in it deserves a second look.
It also explains a pattern that puzzles site owners. Sites that once passed comfortably began failing on responsiveness without anything changing on the site itself. Nothing got slower; the measurement simply started counting the part that was always slow.
Where FID went wrong
Its optimism was the fundamental flaw. Measuring only the initial wait meant heavy click handlers, slow rendering after an interaction and sluggish interfaces went entirely unrecorded, so a page could hold a good score while feeling unpleasant to use.
Being limited to the first interaction narrowed it further. Delays that built up later in a visit — as more scripts loaded, as a chat widget woke up, as a form grew more complex — never appeared. And because it needed a genuine interaction, coverage was uneven: pages that people scroll and leave contributed nothing, so the reported figure often described a small and unrepresentative slice of visits.
What to do about it
Stop reporting it, and stop asking developers to improve it. Anything built around the old metric should be re-measured against Interaction to Next Paint before you decide it is fixed, because work that satisfied FID may do nothing for the successor. Update any dashboard, template or client report that still carries the column.
The underlying advice has not changed much, only widened. Reduce the amount of JavaScript running on the page, audit third-party tags rather than adding to them, break long tasks so the browser can respond between them, and test on a mid-range phone with processor throttling rather than on a fast laptop. Judge the result on field data from real visits, and read it beside the current Core Web Vitals rather than in isolation, as part of ordinary page speed work.