Descriptive link text: 305 of 5,272 homepage links have no text at all

FIELD TEST · 2026-08-28 · 27 homepages · single crawl · 5,272 anchors parsed

Sample and method: the same 30-site panel in use since 2026-08-15, each homepage fetched once on 2026-08-28 at 11:37 (+0800) with a browser user agent and no JavaScript execution. For every <a> element we recorded the text between the tags, whether it wrapped an image, the image's alt length, and whether the anchor carried an aria-label or a title.

Advice about descriptive link text usually attacks the wrong problem. On 27 homepages, only 26 of 5,272 links use a generic phrase like "learn more" or "see more". But 305 links carry no text at all, and for 237 of those, the page hands a crawler nothing that Google's documentation says it will read.

How we measured it

One HTTP request per site, no rendering, no retries. The inclusion rule was fixed before the run: final status 200, decoded body of at least 10,000 bytes, body containing <body. stackoverflow.com and medium.com returned 403 and www.reddit.com returned an 8,393-byte shell, leaving 27 sites.

Link text means the character data between <a> and </a>, whitespace collapsed, including text inside nested elements. Anchors with nothing left after that were sorted into one bucket each, in this order: image with non-empty alt, then title attribute, then aria-label, then image with an empty alt, then nothing. Generic text was matched against a fixed list of 20 English phrases written before the run.

What Google reads when there is no descriptive link text

Two things, and aria-label is not one of them. Make your links crawlable, retrieved 2026-08-28, names exactly two fallbacks: "As a fallback, Google can use the title attribute as anchor text if the <a> element is for some reason empty", and "For images as links, Google uses the alt attribute of the img element as anchor text". The same page marks <a href="/add-to-cart.html"><img src="…" alt=""/></a> as bad, under the heading "empty alt text and empty link text".

Sorted against that rule, the 305 look like this:

So 68 of 305 empty links are covered by something the documentation describes. The largest single group, 102 anchors, leans on an attribute that page never mentions — an accessibility fix that happens to be the most common way a link on this panel is named.

Where the 305 sit

Not evenly. Six of the 27 sites — vercel.com, linear.app, www.cloudflare.com, www.netlify.com, www.wired.com and www.wikipedia.org — have zero anchors without text. Six others hold two thirds of the total.

The shape differs by site. On arstechnica.com most of them wrap an article thumbnail, so the alt can carry the headline. On news.ycombinator.com they are the small vote and comment controls, which have no destination worth describing.

Generic link text is rare; missing text is not

Twenty-six anchors in 5,272 matched the generic list. Seventeen of the 27 sites had none.

The concentration is the interesting part: seven of the 26 are on techcrunch.com, six on supabase.com, five on railway.com. Meanwhile the median link on this panel carries 11 characters of text, and 334 anchors carry 100 characters or more — long link text is 13 times more common here than generic link text.

What this means for you

Run the count on your own homepage before you rewrite a single "learn more". The ratio here says the cheap win is elsewhere.

1. Count anchors with no text. If a link wraps an image, give that image an alt — Google's page says it uses that as the anchor text, which turns a nameless link into a named one with no visible change to the page.
2. Check whether your icon links rely on aria-label alone. That attribute is right for screen readers and it stays. The documented fallback for search is title or the image's alt, so a link that has only aria-label is described for one audience and not the other.
3. Apply the out-of-context test to what is left. Google's wording: "Try reading only the anchor text (out of context) and check if it's specific enough to make sense by itself."

The method for step three, including what to do with the 100-character links, is in how to write anchor text. If the question underneath all of this is whether AI answers are quoting your pages at all, that starts one layer up, at how QueryWin works.

What this run cannot tell you

Whether the 58 alt texts are any good. We measured that the alt is not empty, not that it describes the destination — a thumbnail alt reading "photo" passes our count and fails the purpose. We also truncated recorded link text at 120 characters, so "334 anchors of 100 characters or more" is a floor with no ceiling attached — we do not know how long the longest one runs. And a fixed list of 20 phrases will miss generic text we did not think of.

One earlier run left a question open here. The nofollow links survey from 2026-08-23 counted 311 links with empty anchor text and said plainly that it had not checked whether the images inside them carried alt. This run checked: 58 did. The two totals are not comparable as a trend — that run counted only links with a usable href, this one counts every anchor element, and they are five days apart.

Common questions

How did you measure this?

One fetch per homepage on 2026-08-28 with a desktop browser user agent, parsing the delivered HTML and collapsing whitespace in each anchor's text. Buckets were assigned in a fixed priority order so every empty-text anchor lands in exactly one of them, which is why the five rows add to 305.

Does aria-label hurt SEO?

Nothing here shows that. The finding is narrower: Google's page on link text names title and image alt as fallbacks and does not name aria-label, so a link with only an aria-label has no documented name for search.

Is an empty alt on a linked image always wrong?

Inside a link, Google's page prints that exact combination as its bad example. On a decorative image that is not a link, an empty alt is the correct choice — the split is covered in missing alt text on this same panel.

How many links should a homepage have?

This measurement does not answer that. It records 27 homepages ranging from 27 to 428 anchors, with no outcome data attached to either end.