Link Building and Backlinks

Link building is the practice of earning hyperlinks from other websites to your own. A backlink acts as an editorial endorsement: a signal that another site considered your content worth referencing. Google has treated backlinks as a ranking signal since its inception, and they remain a powerful lever in competitive SEO.

Why do links still matter?

Google’s original PageRank algorithm was built on the premise that a page linked to by many other pages is likely more valuable than one that isn’t. Decades of algorithm updates have made that calculus far more nuanced, but the underlying principle holds.

Links transfer authority, establish topical relevance, and signal that a domain is recognised within its field. A single high-quality link from a relevant, authoritative publication can move rankings more than dozens of low-quality directory submissions. The gap between a good link and a bad one is enormous.

How much weight they carry is contested, and worth seeing plainly rather than assuming. Google’s own position has moved. In 2016, search quality strategist Andrey Lipattsev named links as one of Google’s top three ranking factors. By 2023, Gary Illyes was saying the opposite at two Pubcon events: in February that “links are important, but not as important as people think”, and in a September Q&A that people overestimate their importance and that links had not been a top-three factor “for some time”, adding that there is no universal top three because the leading signals differ from site to site.¹ Both remarks come from attendees’ posts on X; neither trade report cites a recording. Duy Nguyen of Google’s search quality team put it similarly in the November 2022 SEO office hours: “backlinks as a signal has a lot less significant impact compared to when Google Search first started out many years ago. We have robust ranking signals, hundreds of them.”²

Google speaks for Google, and a self-report is not independent evidence. SEOs competing on link-heavy commercial queries frequently read their own data differently, and Google has an obvious interest in discouraging link acquisition. Both things can be held at once: links remain a signal worth investing in, and the claim that they are Google’s most heavily weighted signal is not one Google supports.

What makes a link valuable?

Several factors determine a link’s value:

Relevance. A link from a site in your niche or covering your topic is worth more than one from an unrelated domain. Google evaluates topical context: both the linking domain’s focus and the specific page the link sits on.

Authority. The strength of the linking domain matters. Links from established, frequently cited publications transfer more authority than links from new or low-traffic sites. Domain Rating (Ahrefs) and Domain Authority (Moz) are imperfect but useful proxies.

Placement. Editorial links inside body content carry more weight than footer links, sidebar widgets, or links in boilerplate sections. A link a human editor chose to include is the kind Google trusts most.

Anchor text. The clickable text of a link sends a relevance signal. Descriptive, natural anchor text is valuable. Exact-match keyword anchors pointing to the same page in volume look manipulative and can trigger algorithmic or manual penalties.

Uniqueness. A link from a domain that hasn’t linked to you before carries more weight than additional links from the same domain.

Dofollow vs nofollow. Standard links pass authority. Links marked rel="nofollow", rel="ugc", or rel="sponsored" are treated as hints to Google about the nature of the link and generally do not pass the same equity. High-traffic nofollow links from major publications still drive referral visits and build brand visibility.

Link acquisition strategies

The strategies below are the main categories. For running them as a repeatable programme, with prospect qualification, outreach sequencing and the conversion rates to expect, work from the link building guide.

Content-led acquisition (link earning)

Durable links come from content people want to reference. Original research, industry surveys, data-driven studies, and definitive reference resources attract editorial citations because they provide something others can’t easily replicate. This requires upfront investment but builds a compounding asset.

Digital PR

Proactively pitching stories, data, or expert commentary to journalists and publications earns editorial links from high-authority news and media domains. Digital PR links tend to be valuable: the same publications that dominate traditional media metrics also dominate link equity transfer.

Guest posting

Guest posting means writing for another site in your field in exchange for a link. It stays effective when the article genuinely serves the host’s audience, and becomes a link scheme when the primary purpose is the link rather than the reader.

Prospecting-led acquisition

The remaining approaches share a shape: find a site that has already demonstrated it will link to something like yours, then give it a specific reason to link to you instead. Broken link building offers your page as a replacement for a dead outbound link. Unlinked mention reclamation turns an existing reference into a hyperlink. Competitor backlink analysis works backwards from the domains already linking to your rivals. All three are prospecting methods rather than distinct strategies, and link prospecting and outreach covers them in full, along with qualifying prospects and writing the outreach that converts them.

Which page you point links at is a strategy decision as much as an acquisition one, and it varies by site type. On a store the answer is settled: category pages, not individual products, are the durable targets, because products go out of stock and get discontinued while the category persists. Ecommerce site architecture covers why the hierarchy makes that work.

Do backlinks help AI search visibility?

Yes, but indirectly, and the indirect route is better evidenced than the direct one.

Measured on their own, links correlate poorly with AI visibility, and the metric that does better is not a link-only metric.

Study	Metric	Relationship with AI visibility
Ahrefs, Dec 2025, 75,000 brands	Backlinks	Weak (0.19–0.24)
Ahrefs, Dec 2025, 75,000 brands	Domain Rating	Weak (0.27–0.33)
Semrush, Oct 2025, 1,000 domains	Authority Score, vs AI mentions	Moderate (0.57)
Semrush, Oct 2025, 1,000 domains	Authority Score, vs AI share of voice	Weak-to-moderate (0.36)
Ahrefs, Dec 2025, 75,000 brands	Branded web mentions	Strong (0.66–0.71)

These are Spearman coefficients: they describe how closely two rankings agree, not how much one thing causes another, on a scale where 1 is an exact match and 0 no relationship at all.³⁴ Two things flatten the Semrush rows. Against AI share of voice, the closer analogue to what Ahrefs measured, Authority Score sits level with Ahrefs’ Domain Rating range rather than above it;⁵ and Authority Score is not a link-only metric, since Semrush document the domain’s own organic traffic as one of its inputs.⁶ The score that already contains a ranking signal is the one that correlates better, which points at rankings doing the predicting. Domain authority explained takes that argument further.

Two indirect mechanisms are firmer than either correlation.

The link graph decides what gets crawled into the training corpus. Common Crawl computes PageRank and Harmonic Centrality across its link graph and uses them to set crawl priority, and says sites with higher centrality “are crawled more frequently”.⁷ Common Crawl is the raw material most LLM pre-training is built from: Mozilla’s review of 47 models published up to October 2023 found at least 64% used a filtered version of it.⁸ Neither fact makes a strategy. Common Crawl is describing the value of its own data, the Mozilla figure predates every current model, and being in a training corpus is not being cited at inference.

Rankings feed retrieval, for a minority of citations. AI answers draw on pages that already rank, so anything that lifts rankings improves the odds of entering the source pool. Ahrefs found 37.9% of AI Overview citations came from URLs in the first ten blocks of the results page, against roughly 76% in its July 2025 study; blocks include ads and features, so this is not the organic top ten. Ahrefs reads the fall as query fan-out, AI Overviews drawing on background searches rather than the page the user would have seen, but the two studies are not like for like: the earlier one counted only the three most visible citations in each answer, and Ahrefs improved its citation parsing in between.⁹ Read the level rather than the trend. Ranking gets a page into the pool for a minority of citations, so a link programme justified mainly by its knock-on effect on AI citations rests on a partial effect.

One finding reconciles the studies. Semrush found nofollow links correlate with AI mentions almost identically to followed ones, 0.509 against 0.504 on Spearman.⁴ If passed authority were the mechanism, followed links would pull clearly ahead. They do not, so the mention is doing the work rather than the equity it carries. A campaign built to earn coverage serves both surfaces; a campaign built to place links serves only one. Digital PR and brand mentions cover the coverage side.

What to avoid

Buying links. Google’s guidelines prohibit paid links that pass PageRank. The usual consequence is not a penalty but nullification: SpamBrain, Google’s “AI-based spam-prevention system”, detects both sites buying links and sites passing them, and Google’s aim is to “nullify these unnatural links at scale”, with the result that “ranking may change as spammy links are neutralized and any credit passed by these unnatural links are lost”.¹⁰ The money buys a link that stops counting rather than a demotion. Manual actions are a separate and rarer outcome for the clearest cases, and those do suppress. Either way the risk-to-reward ratio has deteriorated as Google’s detection has improved.

Link exchanges. “Link to me, I’ll link to you” schemes are explicitly called out in Google’s spam policies. Occasional reciprocal links from genuinely relevant partners are normal; systematic exchanges are not.

Private blog networks (PBNs). Networks of sites created specifically to pass links violate Google’s guidelines. Detection and devaluation of PBN links is ongoing and improving.

Low-quality directories. Submissions to generic, low-quality aggregator sites provide no meaningful authority and may dilute your link profile.

How do you measure link building success?

Track progress through a combination of:

• New referring domains (growth in unique linking sites, not total link count)
• Link quality (Domain Rating or Domain Authority of acquiring domains)
• Organic ranking movement for target pages
• Organic traffic growth to link-targeted pages over time

Links take time to influence rankings: weeks to months, depending on crawl frequency and the authority of the linking domain. Short-term attribution is unreliable; measure link building impact over quarters, not days.

The same compounding logic applies to both international link building and local link building in each target market. Each geography requires its own link profile built from scratch, following an independent SEO timeline. Link prospecting and outreach are the discovery mechanisms in every context. Black-hat SEO techniques, including paid links and link schemes, carry the same penalty risk across all markets and should be avoided in favour of white-hat SEO acquisition that earns links through genuine editorial value.

Footnotes

1. Links are not a top 3 Google Search ranking factor, says Gary Illyes — Search Engine Land, 25 September 2023. From a Q&A at Pubcon Pro Austin, 21 September 2023; no trade report cites a recording, and the wording survives as attendees’ posts on X collected by Search Engine Roundtable. The February 2023 quote comes from the Pubcon keynote Q&A, likewise via attendees’ posts (Search Engine Roundtable). ↩

2. November 2022 Google SEO office hours transcript — Google Search Central ↩

3. Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) — Ahrefs. Spearman correlations, December 2025. The authors explicitly caution that correlation is not causation. ↩

4. Do Backlinks Still Matter in AI Search? Insights from 1,000 Domains — Semrush. 1,000 randomly selected domains across ChatGPT, ChatGPT Search, Gemini, AI Overviews and Perplexity, October 2025. Pearson and Spearman correlations. ↩ ↩²

5. Do Backlinks Still Matter in AI Search? Insights from 1,000 Domains — Semrush. Authority Score is reported against two outcomes: AI mentions (0.65 Pearson / 0.57 Spearman) and AI Share of Voice (0.23 Pearson / 0.36 Spearman). ↩

6. What is Authority Score? — Semrush. Documents the three facets: link power, organic traffic, and spam factors. ↩

7. How SEOs Are Using Common Crawl’s Web Graph Data for AI Ranking Signals — Common Crawl. Common Crawl describing its own Web Graph release. ↩

8. Training Data for the Price of a Sandwich: Common Crawl’s Impact on Generative AI — Mozilla Foundation. Review of 47 text-generation LLMs published between 2019 and October 2023. ↩

9. Update: 38% of AI Overview Citations Pull From The Top 10 — Ahrefs. 863,000 keyword SERPs and 4 million AI Overview URLs, March 2026. ↩

10. December 2022 link spam update — Google Search Central Blog ↩