X-UA-Compatible on 27 homepages: 2 still ship the meta tag, 0 send the header

FIELD TEST · 2026-09-06 · 27 homepages · one request each · a tag for a retired browser

Sample and method: 30 homepages fetched once each on 2026-09-06 with a desktop Chrome user agent (Chrome/128, macOS), redirects followed, no JavaScript executed, exit node in Japan. Three dropped out — stackoverflow.com, medium.com and www.reddit.com each returned 403 — leaving 27.

Two of 27 homepages still carry an X-UA-Compatible meta tag. Zero send it as a response header. MDN says user agents now ignore the pragma, and the browser it was written for ended desktop support in 2022. The two sites that kept it do not send the same value, and only one of them sends the value the documentation requires.

How we measured this

One GET per site, desktop Chrome user agent rather than Googlebot, redirects followed, raw HTML written to disk before anything was parsed. Inside each head we listed every meta element carrying an http-equiv attribute, read its value verbatim, and counted the elements that came before it. Response headers were lowercased and checked for the same name. The table below counts declarations; site counts are given separately wherever the two differ.

# The same reading on one of your own pages
curl -sL -A 'Mozilla/5.0' https://example.com/ \
  | grep -io '<meta[^>]*http-equiv[^>]*>'

# And the header form, which nobody on this panel sent
curl -sI -A 'Mozilla/5.0' https://example.com/ | grep -i x-ua-compatible

Four things this measurement cannot do, stated here rather than in a footnote:

1. Homepage only, one request each. A homepage is not a site.
2. JavaScript was not executed, so a tag written in at runtime would not be counted. For a meta http-equiv declaration that risk is small. It is not zero.
3. We did not test any browser behaviour. Nothing here shows what a browser does with these tags, only what was sent.
4. We cannot tell whether a missing tag was a decision or was never there in the first place. One response will never separate those two.

Nine http-equiv declarations, on four of 27 sites

Across all 27 homepages there are nine http-equiv meta declarations in total, and they sit on four sites. The other 23 declare none at all.

MDN documents seven values for the attribute: content-language, content-type, content-security-policy, default-style, refresh, set-cookie and x-ua-compatible (meta http-equiv, MDN, read 2026-09-06). GitHub's four x-pjax-* names are not among them, and each of those four carries a 64-character hex digest as its content. Four of the nine declarations on this panel are outside the documented set.

One row in that table changed after we first wrote it. The initial pass counted slack.com's refresh alongside the rest as a plain declaration. Re-reading the raw HTML showed it sits inside a noscript block, which makes it a fallback for visitors with no JavaScript rather than something every visitor gets. The table already carries the correction, and it changes how that line should be read: not "this homepage redirects" but "this homepage has a plan for the case where nothing runs".

Only one of the two x-ua-compatible values is IE=edge

MDN is specific about the value: "If specified, the content attribute must have the value IE=edge." One of the two sites sends exactly that. The other sends IE=edge,chrome=1.

The chrome=1 half opts into Chrome Frame, which the Chromium blog introduced as "a secure plug-in that brings a modern engine to old versions of Internet Explorer" and then discontinued: "Given these factors we've decided to retire Chrome Frame, and will cease support and updates for the product in January 2014." (Retiring Chrome Frame, Chromium Blog, 13 June 2013, read 2026-09-06.) So one homepage on this panel still tells a discontinued browser to reach for a plug-in whose support ended in January 2014.

The two positions are as far apart as the two values. On www.wired.com the tag is the third element inside the head, preceded only by the title and one other meta. On slack.com it is the 48th, with 18 link elements, 10 script elements, 17 other meta tags, a title and a noscript all ahead of it.

Those positions are reported as measurements, not as a pass or a fail. Older Microsoft guidance said this tag had to appear near the top of the head, and we could not find a currently published Microsoft page stating that requirement in those words. So we counted the elements and stopped.

The x-ua-compatible header: zero of 27

Not one of the 27 responses carried X-UA-Compatible as an HTTP response header. Both sites that declare it at all do so in the markup only.

Microsoft's specification treats the two forms as one instruction with a defined priority: "If both of these instructions are sent, the developer's preference (meta element) takes precedence over the web server setting (HTTP header)." The same page defines the values — IE=edge is "The highest supported document mode of the browser", IE=5 is Quirks Mode, IE=EmulateIE7 gives IE7 mode when a valid doctype is present and Quirks Mode otherwise — and then says of the feature as a whole: "This functionality will not be implemented in any version of Microsoft Edge." (X-UA-Compatibility Meta Tag and HTTP Response Header, Microsoft Learn, read 2026-09-06.)

Two homepages are still addressing a browser that stopped shipping, in a syntax its successor never implemented.

The lifecycle page closes that loop: "Internet Explorer (IE) 11 is the last major version of Internet Explorer. On June 15, 2022, the Internet Explorer 11 desktop application ended support on certain versions of Windows 10." (Internet Explorer 11 end of support, Microsoft Learn, read 2026-09-06.) Internet Explorer mode in Edge is a different thing and, per the same page, "will be supported through at least 2029". We measured nothing about it.

The other piece of legacy head furniture

Same 27 homepages, one more old habit. <meta name="keywords"> is on two of them, www.nytimes.com and www.wired.com. Two of 27 for that one, two of 27 for a tag addressed to a browser that stopped shipping, and www.wired.com is on both lists.

That overlap is the closest thing to a pattern here, and it is one site, so we are not building a theory on it. Same panel size, same method, one tag at a time: the meta keywords tag survey counted that tag, and the charset meta tag survey counted the encoding declaration in the same head.

What this means for you

There is no search dimension in this. The tag does not affect crawling or indexing, and this data set contains no ranking information. What it gives you is a maintenance reading: a line in your head section addressed to software that no longer exists usually means nobody has opened that template in a long time.

1. Read what your own page sends before deciding anything. The two commands above settle it in a few seconds.
2. If the tag is there and the value is not IE=edge, you have inherited someone's copy-paste. chrome=1 in particular asks for a plug-in whose support ended in January 2014.
3. Do not add this tag to a new page. MDN says user agents now ignore it, and Microsoft says the functionality will not be implemented in any version of Edge.
4. Before deleting it, check who your users actually are. We did not test Internet Explorer mode in Edge, which Microsoft says is supported through at least 2029, and a measurement of 27 public homepages says nothing about an internal application.

The rest of that head section is still doing work even when this tag is not, and you can see everything an AI crawler receives from a page in one pass with the AI crawler accessibility check.

Common questions

How did you measure this?

One GET per homepage on 2026-09-06, 30 attempted and 27 included, desktop Chrome user agent, redirects followed, no JavaScript executed, exit node in Japan. Every meta element with an http-equiv attribute inside the head was listed, its value read verbatim, and the elements before it counted. Response headers were checked for the same name. Three sites returned 403 and were dropped.

Can I delete the x-ua-compatible meta tag?

Nothing in this measurement argues for keeping it. 25 of the 27 homepages here do not have it, MDN says user agents now ignore it, and Microsoft says the functionality will not be implemented in any version of Edge. We tested no browsers ourselves, so if part of your audience arrives through Internet Explorer mode in Edge, check there before you delete.

Does the x-ua-compatible meta tag affect SEO?

Nothing here shows that it does. This measurement contains no ranking data, and the tag is an instruction to one browser family about document modes rather than something a crawler acts on.

What does IE=edge actually mean?

Microsoft's value table defines it as "The highest supported document mode of the browser". The same table gives IE=5 as Quirks Mode and IE=EmulateIE7 as IE7 mode when a valid doctype is present, Quirks Mode otherwise. Those definitions describe Internet Explorer, whose desktop application ended support on June 15, 2022.

Should the tag be first in the head?

We could not answer that one. Older Microsoft guidance asked for it near the top of the head, and we found no currently published Microsoft page stating that requirement in those words. So the two positions above are counts of elements, and we made no judgement about either.