- Google's published test for a bad link is intent, not quality: a link created "primarily for the purpose of manipulating search rankings" is link spam, whatever the site's metrics look like.
- That means a link from a weak but genuine site can be fine, and a link from a strong site can be spam. Authority scores do not answer the question.
- Check in order: would this link exist without SEO, is the page it sits on real, is it relevant to what you do, and only then look at metrics.
Most guides answer this with adjectives. Good links are "authoritative" and "relevant", bad links are "spammy" and "low quality". That is fine until you are looking at an actual link and have to decide.
Google publishes a sharper answer, and almost nobody quotes it.
The one test that decides it
A link is bad when it was created primarily to manipulate search rankings. That is Google's own wording in its spam policies, and it is a test of intent rather than a measure of quality. The same sentence covers buying links, excessive link exchanges, automated link creation and optimised anchor text in guest posts, because the common factor in all of them is why the link exists.
This reframes the question you are asking. You are not grading a website. You are asking whether the link would exist if search engines did not.
If a real editor put it there because it helped their reader, it is a good link, even from a small site. If it exists because money changed hands, or because a script placed it, or because you agreed to link back, it is spam by Google's definition, however impressive the domain looks. Intent is the line, and the comparison below is what that line looks like in practice.
Good vs bad, side by side
The clearest way to see the difference is by what caused the link, not by what the link looks like. Every row below is the same signal read two ways.
| Signal | Good link | Bad link |
|---|---|---|
| Why it exists | A writer needed a source and picked yours | You asked, paid, swapped or automated it |
| Who chose the anchor text | The publisher, so it reads naturally | You did, so it is your target keyword |
| The page it sits on | Has its own audience and gets traffic | Exists to hold links and nothing else |
| Editorial control | The publisher can remove it and would defend it | You control the placement, or you rented it |
| Topic | Connected to what you actually do | Unrelated, or the site covers everything |
| The company it keeps | Sits among links you would be happy to be listed with | Surrounded by casinos, loans and pharmacy links |
| What it does for you | Sends occasional real visitors | Has never sent a human being |
Read down the "why it exists" row first. Most links that fail do so on that row alone, and the rest of the table is confirmation rather than the decision.
How to check a link, in order
Check a link in the order that kills it fastest, so you stop as soon as you have your answer. Working top down means most decisions take under a minute.
- Would this link exist if search engines did not? If the honest answer is no, stop. Nothing further matters, and this is the question Google's policy actually asks.
- Is the page it sits on a real page? Open it. Does it have an author, a date, a point of view and other content around it? A page that exists only to hold outbound links fails here.
- Does the site have an audience? Look for comments, shares, a newsletter, anything showing humans arrive. A site with no readers passes nothing on, whatever its metrics say.
- Is the anchor text something a human would write? Your exact commercial keyword as anchor text on a site you do not control is the single most common tell.
- Is it relevant to what you do? Not the same industry necessarily, but a reason for the connection that you could explain to someone in one sentence.
- Only now, look at the metrics. They are useful for prioritising between links that already passed, and useless for deciding whether to pass at all.
Step six is the one most people do first, and doing it first is where the trouble starts.
Why domain authority is the wrong first question
Domain authority is a third-party estimate, not a Google metric, and it describes a whole domain rather than the page linking to you. Moz, Ahrefs and Semrush each calculate their own version, the numbers disagree with each other, and none of them is a number Google publishes or uses.
The practical problem is the unit. A domain-level score tells you about a website. Your link sits on one page of that website, and a strong domain routinely carries pages that nobody reads, nobody links to and Google barely crawls. A link from a forgotten page on a famous site can be worth less than a link from a well-read page on a small one.
This is why the metric is last on the list rather than absent from it. Once several links have passed the intent test, a score is a reasonable way to sort them. As a first filter it will approve paid links on high-scoring domains and reject genuine editorial links on small ones, which is exactly backwards. Seeing that pattern in real link profiles is what the examples below are drawn from.
What bad links look like in practice
Google names the specific practices rather than leaving it to judgement, and the list is more concrete than most summaries of it. Its spam policies call out buying or selling links for ranking purposes, including exchanging goods, services or a free product for a link. Excessive link exchanges and partner pages that exist only for cross-linking are on the list. So are automated link-building programs.
Several entries catch people who think they are doing something legitimate. Advertorials and native advertising where payment is involved and the links pass ranking credit are link spam unless those links are qualified. Optimised anchor text in guest posts and press releases is named directly. Low-quality directory and bookmark links, links embedded in widgets distributed across sites, and links pushed into the footers or templates of many sites all appear on it.
The one that surprises people most: requiring a link as part of a contract or terms of service, without letting the other party qualify it, is on the list too. If your agency agreement, your theme licence or your supplier contract mandates a link, that is the practice Google is describing. Once you can recognise these, the question becomes what to do about the ones already pointing at you.
What to do about bad links you already have
Most bad links pointing at a site need no action at all, because Google already ignores the majority of them. Anyone can link to you, so penalising sites for links they did not create would make negative SEO trivial. Google's systems are built on that assumption and discount the obvious spam rather than counting it against you.
Action becomes worth taking in two situations. The first is a manual action in Search Console, which names the problem and is the one case where the disavow tool genuinely applies. The second is when you built the links yourself, or paid someone who did, and they are still live.
Removal beats disavowal when removal is possible, because a link that no longer exists cannot be assessed at all. Where the site will not remove it, the disavow file exists. Use it narrowly and deliberately, not as routine maintenance, and keep a record of what you disavowed and why so a future audit does not have to guess.
FAQ
Do nofollow links count as bad links?
No. A nofollow link is a link the publisher has chosen not to pass ranking credit through, which is a labelling decision rather than a quality judgement. Nofollow links from real publications still send visitors, still build recognition, and still appear in a healthy link profile.
Can a single bad backlink hurt my rankings?
A single link almost never causes a problem on its own, because Google discounts links it does not trust rather than punishing the target. Patterns cause problems: many paid links, repeated exact-match anchors, or a profile made mostly of links you created.
How many backlinks should a site have?
There is no target number, and chasing one is how link profiles end up full of the links on the right-hand side of the table above. A smaller profile of links that pass the intent test outperforms a larger profile that does not.
Is buying a link ever safe?
Buying a link for ranking purposes is against Google's published policy, without a threshold or an exception for high-quality sites. Paying for advertising is fine when those links are qualified so they do not pass ranking credit, which is what the qualification attributes exist for.
Check your own links
Two free things on this site do the work that follows from this page. The Backlink Gap Finder shows which sites link to competitors and not to you, which is where genuine link opportunities usually hide. The Competitor Backlink Scanner is a Claude skill that runs the same analysis in more depth.
If you are looking for places to earn links rather than judge them, the free backlinks list is a vetted set of legitimate sources, and digital PR for link building covers earning editorial coverage at scale. For what link building achieves once AI assistants are the ones answering, see link building in 2026.
Sources
- Google Search Central, Spam policies for Google web search, link spam section. Read 23 August 2026.