A backlink is a hyperlink on a third-party site that points to your site. That’s the entire definition — but the second-order effects of those links are what make backlinks the most consequential signal in SEO. Google’s original 1998 PageRank algorithm was built around link math, and twenty-eight years later, in spite of every algorithmic update and AI overhaul, link signals still decide who ranks for competitive queries and who gets cited by AI engines.

“Backlinks are votes you don’t cast for yourself. Every other ranking signal is something you control on your own page; backlinks are the only signal that depends on what the rest of the open web thinks of you. That’s exactly what makes them the hardest to fake — and the most expensive to lose.”

What a backlink actually is

Mechanically, a backlink is an HTML <a href="..."> element on a page that isn’t yours, where the href resolves to a URL on your site. Google’s crawler follows the link from the source page to your destination page, records the source URL, the destination URL, the visible anchor text, the surrounding context, and the link attributes (rel= values like nofollow, sponsored, ugc). All of those data points become part of the link graph — Google’s map of who links to whom across the entire web.

The link graph is the input to PageRank and to every modern descendant of PageRank. It’s also the input that AI engines read when they need to rank candidate sources for citation. When Perplexity or Claude generates an answer with citations, the model isn’t just looking for relevant text — it’s looking for relevant text on pages that the link graph says are authoritative. Authoritative, in this context, means “has more incoming links from pages that themselves have more incoming links.” The math is recursive, which is why it’s hard to game.

The four signals Google reads from every backlink

Not all backlinks are equal — Google extracts at least four distinct signals from each one, and the weighting between them has shifted over time.

The two signals SEO teams routinely under-weight are relevance and context. A link from a high-authority but topically irrelevant source contributes much less than the raw authority number suggests. A link buried in a sidebar of 200 other links contributes far less than the same source linking from inside a written article. When you’re evaluating a link opportunity, those two filters often matter more than the source’s headline domain rating.

Dofollow vs nofollow vs sponsored vs ugc

The link rel= attribute tells Google how to treat the link. Four values matter:

• rel="dofollow" (or no rel at all): the default. Pass full authority and anchor signal to the destination.
• rel="nofollow": originally introduced in 2005 to fight comment spam. Since 2019, treated as a hint — Google may still pass partial signal, especially from high-authority editorial sources.
• rel="sponsored": indicates a paid link (advertorial, affiliate, sponsored content). Google won’t penalize the link, but won’t pass full authority either.
• rel="ugc": user-generated content (forum posts, blog comments, public Q&A). Treated similarly to nofollow.

A healthy backlink profile contains all four — never zero of any. A profile with 100% dofollow editorial links is statistically impossible in the wild and signals manipulation. The right shape: ~70–85% dofollow, ~15–25% nofollow / sponsored / ugc combined, distributed across a long tail of source domains.

Quality over quantity — the math that decides

The single most repeated mistake in link building is chasing referring-domain counts as a vanity metric. Ten links from TechCrunch, The Wall Street Journal, and Wired outperform a thousand links from low-DR PBNs and link farms. The asymmetry is structural: Google’s scoring is recursive, so a link from a page that itself has many high-authority inbound links passes a multiple of the equity of a link from a thin page.

Practical benchmark: when you’re evaluating a backlink opportunity, look at three numbers — the source domain’s referring-domain count, the source page’s referring-domain count, and the source domain’s topical relevance to your industry. If the source domain has 10,000 referring domains but the actual page you’ll be linked from has 0, the link is worth significantly less than the headline number suggests.

How AI engines treat backlinks

AI engines — ChatGPT, Perplexity, Claude, Gemini, and Google’s AI Overviews — don’t rank pages on a SERP the way Google does. They generate answers and pick a small number of pages to cite as sources. The selection criterion isn’t identical to Google’s ranking algorithm, but link authority is one of the heaviest inputs.

The practical effect: in the AI era, citations are zero-sum. If a query has space for three citations and your page is the fourth-most-authoritative answer, you don’t get cited at all. Backlink authority is the dominant filter for breaking into citation slots. We’ve watched dozens of client domains pass the “citation threshold” the moment their referring-domain count crossed the median for their query category — and watched competitors fall out of citation when their link profile stagnated.

Link earning vs link building

The vocabulary matters. Link building is the active process of acquiring links — outreach, guest posts, broken-link replacement, resource-page placements, digital PR. Link earning is the passive process of producing content (or data, or tools, or stories) so good that other sites link to it without being asked.

The 2026 reality: pure link earning is the cleanest profile but rarely scales fast enough on its own. Pure link building is fast but trips spam filters and is increasingly unsustainable. The combination — a constant stream of link-worthy assets supported by disciplined outreach — is what actually moves authority. That sequence is the subject of the Link Building Strategies article in this cluster.

Common mistakes

• Chasing volume over quality. 1,000 low-DR links rarely beat 10 from authoritative editorial sources.
• Ignoring relevance. A high-authority backlink from an off-topic source contributes far less than its DR suggests.
• Same anchor text on dozens of inbound links. The fastest way to trip Penguin / spam filters. Diversity is the rule.
• Building only to the homepage. Deep-page links matter for the pages that actually rank. Most ranking pages are not the homepage.
• Ignoring nofollow opportunities. A profile with 100% dofollow looks artificial; nofollow contributes to authenticity and traffic.
• Not auditing the existing profile. Most sites have at least 5–15% toxic or low-value links that should be addressed before chasing new ones. Covered in the Toxic Backlinks article.

Where to start

Before any new link work, the order of operations is: (1) audit the existing profile, (2) understand the link gap to the top-ranking competitors for your target queries, (3) prioritize a small number of high-relevance, high-authority opportunities, (4) build the content asset that earns those links, (5) execute outreach. Skipping the audit is the most common reason link campaigns stall.

The next chapter — Domain Authority — covers the metrics SEO teams use to score the link graph (DR, DA, link equity), what they actually predict, and where they mislead.