Iframes without titles

It gives the frame an accessible name, which is what a screen reader announces when reaching it.

It appears in the list of regions assistive technology offers for navigation, so the frame can be skipped or entered deliberately.

It describes the frame's purpose rather than its technology: what is inside, not that it is an embed.

It is separate from any title inside the embedded document, which a user cannot see from outside.

One short phrase is enough, and it is the whole of the fix.

One unnamed frame is a minor annoyance: the user enters it and finds out.

Four unnamed frames are four identical entries in the region list, with no way to choose between them.

Advertising and tracking frames add to the count without adding anything a user wants, which makes the list longer and less useful.

A page with a video, a map and two advertisements presents four regions, three of which the user does not want and cannot identify.

Which is why a page with many frames is where this stops being cosmetic.

A frame used purely as a transport for a script, carrying no content — those should be hidden from the accessibility tree rather than titled.

A tracking pixel implemented as a frame, which has nothing to announce.

A frame with an empty body kept for a legacy reason.

For each of those, marking it hidden from assistive technology is correct and is safe precisely because there is nothing focusable inside.

So the decision per frame is between a title and being hidden, and a frame with content always takes the title.

Most frames on a typical page are inserted by third-party scripts, and the attribute is not yours to set in the markup.

Many providers offer a configuration option for the title, which is worth looking for before concluding it is impossible.

Where none exists, the attribute can be set after insertion, though that is fragile against the provider re-rendering.

Where the frame carries no content, hiding it from the accessibility tree is the better answer and is usually achievable.

It is reasonable to report the third-party frames separately from your own, since the fixes differ.

Search the served markup for frame elements and read the title attribute on each.

Check the live DOM too, since most frames are inserted by script and are absent from the served HTML.

Open the accessibility tree in developer tools and look at how each frame is named.

Use a screen reader's region list on a page with several frames, which is the clearest demonstration.

Count how many of the frames a user would actually want to enter, which is usually fewer than the number present.

`accessibility.iframe_title` reports embedded frames without a title, which is this problem directly.

`accessibility.semantic_landmarks` reports the landmark structure the frames sit alongside in the region list.

`accessibility.aria_hidden` reports elements hidden from assistive technology, which is the correct treatment for a content-free frame.

`privacy.youtube_embeds`, `privacy.google_maps` and `privacy.recaptcha` identify the common third-party frames, which is what separates yours from theirs.

Frames inserted by script after load are outside the served markup and are not observed.

The content, not the container: the name of the video rather than the word video.

Enough to distinguish it from the other frames on the same page.

Short — it is announced, so a sentence is worse than a phrase.

Not a duplicate of a heading immediately before it, which would be announced twice in a row.

Not the word iframe or embed, which describes the mechanism and tells the user nothing.

Each frame is a separate document with its own scripts, styles and network connections, so a page with four frames is loading four pages.

A frame contacts its provider on page load whether or not the visitor interacts with it, which is a data transfer to declare as well as bytes to pay for.

Loading a frame only when it approaches the viewport removes both costs for the ones below the fold, and the lazy attribute works on frames as it does on images.

A facade — a static image that swaps for the real frame on click — removes the cost entirely until someone wants it, which suits video and map embeds.

Which means the frames worth titling and the frames worth removing are frequently the same list read twice.