nofollow links across 27 homepages: 61 out of 5,104, and sponsored is used once

FIELD TEST · 2026-08-23 · 27 homepages · single crawl · links parsed

Sample and method: the same 30-site panel in use since 2026-08-15, each homepage fetched once on 2026-08-23 with a browser user agent. We parsed every <a> element in the delivered HTML, resolved each href against the final URL, and recorded the rel and target attributes and whether the link had visible text.

Across 5,104 links on 27 homepages, 61 are nofollow links — 1.2% of the total, and 6.3% of the 963 that point off-site. Nineteen of the 27 sites use no link qualification at all. The two attributes Google introduced to replace generic nofollow are rarer still: one sponsored link in the whole panel, and one ugc.

How we measured it

One HTTP request per site, no rendering, no retries. The inclusion rule was fixed before the run: status 200, body at least 10,000 bytes, body containing <body. Three of the 30 failed it — stackoverflow.com and medium.com returned 403, www.reddit.com returned an 8,393-byte shell — leaving 27, the same three exclusions as the previous three batches.

A link counted if it was an <a> element with an href that was not empty and did not start with #, javascript:, mailto: or tel:. Internal means the resolved host equalled the final host after redirects, so a link from www.shopify.com to a Shopify subdomain counts as external here. That is a measurement choice, not a claim about how Google treats it.

Which sites use nofollow links at all

Eight of 27. The list is short enough to print in full, and its shape is more interesting than the total: the publishers are on it and almost none of the software companies are.

Four of the eight are news or publishing sites, which is where paid placement and syndicated links actually occur. TechCrunch marks 8 of its 9 external links, the highest rate in the panel by a wide margin. The New York Times row is the odd one: its single nofollow is on an internal link, not an external one, which is the case Google's documentation steers away from — for links within your own site it names the robots.txt disallow rule instead.

Twenty of the 27 sites send external links with no qualification at all. That is the default, not an oversight.

sponsored and ugc are effectively unused

Google introduced rel="sponsored" and rel="ugc" as more specific replacements for the blanket nofollow, and its documentation still says sponsored is preferred for paid links. In 5,104 links we found one of each: a single sponsored link on www.theverge.com and a single ugc link on substack.com.

That is not evidence the attributes do nothing. It is evidence that homepages are the wrong place to look for them — paid placements live on article pages and user content lives in comment threads, neither of which is a homepage. What the panel does show is that the migration Google recommended has not reached the front door of any of these sites.

What 5,104 links look like when you sort them

The link count per homepage ranges from 21 to 395, with a median of 165. Internal links are 4,141 of the total, or 81%.

One site breaks the internal-versus-external split entirely. www.wikipedia.org is the language portal, and all 374 of its links are external, reaching 361 distinct hosts — every language edition is a different subdomain. Removing it drops the panel's external share from 18.9% to about 11%, which is worth knowing before anyone quotes the first number as typical.

The 311 links with nothing to read

Six percent of the links in this panel have no visible text. They are icons, logos and image-only tiles, and Google's documentation is explicit about what happens next: for image links it uses the alt attribute as the anchor text, and an empty alt on an empty link leaves the crawler with a destination and no description of it.

The concentration is uneven. arstechnica.com alone accounts for 67 of the 311, then stripe.com and news.ycombinator.com with 31 each. We did not check whether the images inside those links carry alt text, so we cannot say how many are genuinely undescribed. That is the one number this crawl should have collected and did not.

What this means for your site

The practical reading of this panel is that link qualification is a small, targeted job, not a site-wide policy. Four things follow from it.

• Leave ordinary external links alone. Google's documentation says no rel attribute is needed for links you expect it to fetch and parse.
• Mark paid placements with sponsored and comment or forum links with ugc. Almost nobody does this on a homepage, which is exactly why doing it where it belongs costs you nothing.
• Do not use nofollow to shape crawling inside your own site. Google names robots.txt for that, and the one internal nofollow we found is an outlier rather than a pattern.
• Do not ship image links with an empty alt. It is the same as shipping a link with no text.

If you want the method rather than the numbers, the hub-and-spoke version of this is written up as an internal linking strategy, including a sheet for tracking which page links to which. For the wider question of whether crawlers can reach those pages at all, the AI crawler accessibility check answers it in one request.

Common questions

How did you measure this?

One homepage request per site on 2026-08-23 with a desktop browser user agent, following redirects, no JavaScript execution. Links were parsed from the delivered HTML only, so anything a framework injects at runtime is not in these counts.

Do nofollow links still pass any signal?

Google's documentation says links carrying these attributes "will generally not be followed", and separately that the linked pages may still be crawled if they are found another way. It does not describe a partial pass-through, and we did not test for one.

Is 165 links on a homepage too many?

Google publishes no limit, and this panel gives no basis for one either — the count is a design decision about navigation density, not a compliance number. We measured what exists; we cannot tell you what any of these sites would gain by trimming.

Why exclude anchors and javascript: links?

Because they do not point at another page. Fragment links are in-page navigation and javascript: hrefs are the pattern Google's own documentation lists as not recommended, so counting them would have inflated every site's total without adding any routes.