"Dofollow" is not a thing, and that matters
There is no rel="dofollow" in the HTML spec. There never was. A link is followed by default; rel="nofollow" is the markup a publisher adds to stop passing signals. "Dofollow" is jargon SEOs invented to name the absence of an attribute.
That sounds pedantic until you watch what it does to people's judgement. A whole outreach industry has organised itself around a binary that only exists in one direction, and it produces a specific bad habit: checking the attribute first and asking whether the page is any good second. Or never.
The right order is the reverse. Look at the page. Then look at the attribute, if you still care.
What actually changed on 1 March 2020
In September 2019 Google announced two new link attributes and a change in how it would treat nofollow. From 10 September 2019 the attributes became hints for ranking purposes, and from 1 March 2020 they became hints for crawling and indexing too.
"Hint" means Google may choose to follow a nofollow link and may choose to count it. Google's stated reasoning was that nofollow had become blunt — huge categories of genuinely useful links were being blanket-nofollowed by CMS defaults, and Google was discarding real information about the web.
Two consequences follow, and both are practical.
First, a nofollow link is no longer worth zero. It's worth an unknown amount between zero and something. Nobody outside Google knows which, and no tool reports it. Anyone who tells you a nofollow link definitely passes value, or definitely doesn't, is guessing with confidence.
Second, nofollow stopped being a reliable way to hide a page from Google. If you need something excluded, use noindex or block it in robots.txt — see noindex for which one applies when.
The three attributes, and when each one applies
Publishers now have three markers rather than one. They can be combined — rel="sponsored nofollow" is valid and common.
| Attribute | What it declares | Where it shows up |
|---|---|---|
| No attribute | An ordinary editorial link. Signals pass normally. | Body copy on most publisher sites, citations, resource pages, genuine editorial mentions. |
rel="nofollow" | "Don't associate us with this link." A hint since March 2020. | Wikipedia, most large forums, comment sections, and the default outbound setting on a lot of big publishers. |
rel="sponsored" | This link was paid for, in money or in kind. | Advertorials, sponsored posts, affiliate links, paid placements. Required disclosure under Google's spam policies. |
rel="ugc" | A user put this here, not the publisher. | Forum posts, comments, community answers, profile fields, reviews. |
The better test: would a human click it?
Here's the replacement for the attribute check. Stop asking whether a link is followed. Ask whether the page it sits on has readers who might become customers.
The test is measurable, which is what makes it better than a vibe. Take the placement, wait 30 to 90 days, then open GA4 and look at the referral channel for that domain. Sessions, engagement, and — if you're tracking properly — leads.
A nofollow link on a page that sends you forty qualified visitors a month is a customer acquisition channel. Whether Google counts it as a ranking signal is genuinely a secondary question at that point; you're getting paid either way. A followed link on a page nobody reads sends nothing, converts nothing, and gives you a number in a backlink tool.
The test also settles most outreach arguments before they start:
- Nofollow on a large industry publication your buyers read. Take it. Immediately. Take it twice.
- Nofollow in a genuine community answer where the question matches your product. Take it. That's the highest-intent traffic on the internet.
- Followed link on a 'general news' site publishing eight posts a day about insurance, casinos and yoga. Refuse. The followed attribute is precisely why it's dangerous — see paid links vs earned links.
- Followed link in a footer, a sidebar or a 'partners' block. Low value at best, a footprint at worst.
- Nofollow in a directory nobody has visited since 2014. Ignore. Not because of the attribute, because of the nobody.
Why an all-dofollow profile is a red flag
This is the part that surprises founders who've been buying links, and it's simple arithmetic about how the web behaves.
A brand that genuinely gets talked about accumulates links from places that nofollow by default — press sites with blanket outbound rules, forums, Reddit threads, LinkedIn posts, YouTube descriptions, comment replies, aggregators, syndication feeds. Those are not accidents; they're what real attention looks like when it hits a CMS with defaults.
So a natural profile is mixed. Heavily mixed. Nobody can hand you the correct ratio, because it varies enormously by industry — a B2B SaaS with an active community will look nothing like a jewellery brand — and any specific percentage you see quoted was made up by someone selling something.
But the direction is unambiguous. A profile where nearly every referring domain passes a followed link, with commercially-worded anchor text, did not happen naturally. It happened because someone specified it on an order form. Google's link spam systems are pattern-matchers, and that is a very legible pattern.
The practical version: if you're auditing an inherited profile, a suspiciously clean followed ratio is a stronger signal of purchased links than any domain metric. Domain authority is a vanity metric — the shape of the profile isn't.
Nofollow links and getting cited by AI assistants
This is the newest reason the attribute matters less than it did, and it's structural rather than speculative.
The rel attribute is a link-graph instruction. It tells a crawler something about how to weight an edge between two pages. It says nothing about whether the text on the page is readable, and text is what answer engines consume.
When an AI assistant summarises "best options for X in India", it's working from pages it has ingested. If your brand is named, described accurately and placed in the right category on a page that gets crawled, that's the input. Whether the link next to your name carried nofollow does not change what the sentence says about you.
Which flips a long-standing outreach preference. A nofollow mention in a widely-read comparison article on a publication that AI systems read is now plausibly worth more than a followed link on a site with no readers and no reason to be quoted. The same logic runs through dofollow-obsessed guest posting versus digital PR — coverage produces mentions, marketplaces produce attributes.
None of this means links stopped mattering. They matter. It means the attribute was never the thing you were actually buying, and in 2026 it's a smaller fraction of the thing than ever.