Link Building Strategies for 2027 in the AI Search Era

8 October 2026 · The Cresia team

Keep the link building that earns mentions from pages people and AI systems already trust, and cut the tactics that only ever produced a number on a dashboard. Links still help decide which pages get retrieved and believed, but they're now one signal among several. For 2027 planning, that means fewer links, chosen harder, and judged by whether your brand shows up in answers.

Key takeaways

Keep link building that earns mentions from pages people and AI systems already trust; cut volume tactics that only move a dashboard number.
Links still shape which pages get retrieved, but they sit beside content structure, entity clarity and brand mentions.
Treat unlinked brand mentions in credible places as a real outcome, not a consolation prize.
Measure link work against citations and mentions in AI answers on a fixed prompt set, not only domain-level link counts.
Audit the link profile you already have before buying anything new. Fixing pages that already hold authority is usually the cheapest win.

What the 2027 keep-or-cut conversation is about

Search Engine Journal has published a webinar announcement, with the SEO vendor Resolve, built around a simple question: which link building tactics should teams keep or cut for 2027, given off-page SEO, rankings and AI visibility. It's a promotion for a live session, so it doesn't hand you a verdict. What it does tell you is where the industry's attention has moved. Link builders are no longer arguing only about whether a tactic works for Google rankings. They're asking whether it still earns anything once an AI answer sits above the results.

That's the right question, and it's worth answering on your own terms before any webinar does it for you.

Here's the practical version. Imagine a B2B software company with a three-person content team and an agency retainer that delivers twenty links a month. Under the old scorecard, the retainer looked fine if the links were on sites with decent authority metrics. Under a scorecard that includes AI visibility, someone has to ask a harder thing: did any of those twenty links make an assistant more likely to name the company when a buyer asks for options? If nobody can say, the retainer is a candidate for the cut list.

Notice what this doesn't say. It doesn't say links are dead. It says the justification for each link has to be stronger than it used to be.

Why links still matter for AEO and GEO

Links matter for answer engines mainly through two doors.

The first is retrieval. Many AI answers are assembled from pages pulled from a search index at the moment of the question. Google's AI features draw on Google's own index, and other assistants use their own crawlers or partner indexes. Check each vendor's documentation for what it says about sources, because the details differ and change. Where retrieval leans on a conventional index, the things that helped a page get indexed and ranked, including links pointing at it, still help it get pulled into the candidate set.

The second door is corroboration. A model or retrieval layer that sees your brand described the same way on several independent, credible pages is more likely to treat that description as safe to repeat. A link is one way those pages point at you. A plain mention of your name next to the right category is another. For answer engines, the second may count nearly as much as the first, and it costs the other site nothing to give.

That shifts the goal. The old aim was a link from an authoritative domain. The 2027 aim is closer to being described accurately, by sources that matter, in the terms your buyers use.

Search Engine Land has reported that ChatGPT and Google's AI answers favour large retailers in shopping responses. Take the headline as a direction rather than a rule: established brands with broad corroboration tend to get named. If you're mid-size, you can't match that by volume. You can match it by being unusually clear and unusually well-cited within a narrow topic.

How AI answer engines pick and cite sources

No one outside the vendors knows the full selection logic, and it changes. What you can observe is a rough pattern that holds across most assistants that browse.

A question comes in. The system breaks it into one or more searches, retrieves a set of pages or passages, and then composes an answer from the passages that fit. Citations, where shown, usually point at the pages whose passages were actually used. Three things tend to decide whether yours is among them:

Whether the page is retrievable at all. It has to be crawlable, indexable and not hidden behind scripts or logins the crawler can't pass. Check your robots rules and the vendor's published crawler names.
Whether a single passage answers the question cleanly. Engines pull passages, not whole pages. A paragraph that states the answer in its first sentence and supports it in the next two is far easier to lift than a point buried in a long narrative.
Whether the source looks safe to repeat. This is where authority, links, brand mentions and a consistent identity do their work.

Search Engine Journal's companion coverage argues that AI search is changing what a website is for rather than replacing it. That fits. Your pages are now both a destination and a source of raw material for answers, and link building feeds the second job as much as the first.

One caution. Different assistants cite differently. Some show many sources, some show few, and some answer without any. A tactic that moves visibility in one may do nothing in another. Test against the engines your buyers use, not a generic idea of AI search.

Which link building tactics to keep, and which to cut

The sorting rule is simple: would this placement still make sense if search engines didn't exist? If a real person would read the page and a real editor chose to include you, keep it. If the only reason the placement exists is that someone sold it, cut it.

The table below applies that rule. It's an opinionated starting point, not a ranking of effectiveness, so adjust it to your category.

Tactic	Keep or cut	Why	Suggested owner
Original data or research that others cite	Keep	Gives editors a reason to reference you and gives engines a quotable fact tied to your name	Content lead with analytics
Digital PR with a real news hook	Keep	Earns mentions from publications that engines treat as credible	PR or comms
Expert contributions and quotes in trade press	Keep	Builds consistent brand and person associations with a topic	Subject experts with SEO support
Reclaiming unlinked brand mentions	Keep	Cheap, fast, and already aligned with how you're described	SEO specialist
Fixing broken or redirected inbound links	Keep	Recovers authority you already earned	Technical SEO
Bulk guest posts on thin sites	Cut	Low editorial bar, little corroboration value, risk if the pattern looks manufactured	Agency or SEO lead
Paid links without disclosure	Cut	Against search engine guidelines and worthless to a reader	SEO lead
Directory and profile spam	Cut	Adds noise, not credibility	SEO specialist
Reciprocal link swaps at scale	Cut	Obvious pattern, no editorial judgment behind it	SEO lead

Two entries deserve a comment. Guest posting isn't dead as a category; a considered article for a publication your buyers read is fine. What goes is the bulk, templated version. And reclaiming mentions is boring, which is exactly why it gets skipped and why it pays.

What to change in your plan this quarter

You don't need a rebuild. You need a reordering. Here's a sequence a small team can run in a quarter.

Audit what you already have. Export your top linked pages and check which still resolve, which redirect, and which have been replaced by newer content. Link equity pointing at a dead URL helps nobody.
Pick five questions you want to be named for. Make them the questions a buyer would actually type into an assistant, like which tools suit a team of a given size, or how to handle a specific compliance problem.
Build or sharpen one page per question. Lead with the answer, support it with a table or short list, name the entity (you, the product, the category) plainly, and date the page.
Point your outreach at those five pages. Concentrate effort. Ten links to one useful page usually beat one link each to ten forgettable ones.
Run a mention-reclamation pass. Search for your brand name and product names, find pages that mention you without linking, and ask politely.
Retire a retainer line. Pick the lowest-value activity on your cut list and stop paying for it. Spend that budget on research or PR.
Align the facts. Make sure product names, descriptions and leadership details are consistent across your site, your profiles and your press material. Engines reward consistency, and inconsistent descriptions get paraphrased inconsistently.

A concrete example: a logistics software firm wants to be named for a question about cutting last-mile delivery failures. It publishes a short original analysis from its own anonymised operations data, which it fully owns and can describe truthfully, then offers it to two trade publications. The result isn't just two links. It's two credible pages that describe the firm in the same terms as its own page. That's the corroboration an answer engine looks for.

How to measure link work against AI visibility

This is the part that's genuinely messy, and Ahrefs has said as much in its piece on measuring AI search ROI. You won't get a clean attribution line from a link to a revenue number. You can get a defensible trend.

Build a fixed prompt set: perhaps thirty to fifty questions across your five target topics, phrased the way buyers phrase them. Run it on a schedule against the assistants that matter to you, and record three things: whether your brand is named, whether your page is cited, and how you're described. Keep the prompt wording constant, because small changes shift results.

Then pair that log with your off-page activity.

What to track	Where it comes from	How often	What a change tells you
Brand named in answers	Fixed prompt set	Weekly or fortnightly	Whether corroboration is reaching the engines
Your pages cited	Fixed prompt set	Weekly or fortnightly	Whether your passages are being retrieved and used
New referring pages and mentions	Link and mention tools	Monthly	Whether outreach is landing
Referral traffic from AI products	Web analytics	Monthly	Whether citations turn into visits
Branded search demand	Search console	Monthly	Whether awareness is building

Two honest limits. AI answers vary from run to run, so one result means little and a trend over several weeks means more. And referral traffic from assistants is often under-reported because of how clicks are passed along, so don't read a low number as proof of low influence.

If you're using a tool for the monitoring, MediaPilot is the Cresia product aimed at media and AI-search visibility, and its page is the place to see what it covers. Whatever you use, check the vendor's own documentation for which assistants and regions it samples.

Measurement also needs clean tagging. If AI referrals land in the wrong channel bucket, the trend line lies to you. A written tracking specification that says how those visits are classified is a small job that saves a lot of arguing later.

What not to do

A few traps are worth naming because they're tempting in a year when everyone feels behind.

Don't chase a magic AI-link metric. There's no agreed score that predicts citation. Be wary of any tool that sells one without showing you how it's derived.
Don't buy placements to look established. Manufactured corroboration is exactly what ranking and retrieval systems are built to discount, and a pattern of paid links can hurt you.
Don't abandon links entirely. The opposite overcorrection, going all-in on mentions and ignoring links, throws away a signal that still feeds the index many engines rely on.
Don't optimise for one assistant. Behaviour changes with model updates. A strategy built on a single engine's quirks ages quickly.
Don't skip the on-site work. Outreach to a page that buries its answer wastes the placement. The page has to be worth citing before it's worth promoting.
Don't report links earned as the outcome. Report what changed in answers and in demand, and keep link counts as an input.

There's also a case for patience. Corroboration builds slowly, and a quarter is a short window. Set expectations with whoever signs the budget, and say plainly that the first results will be directional.

If your team spans content, PR and analytics, who owns what is half the battle. Growth teams often end up holding the whole thing by default; the growth teams solution page describes how Cresia frames that work.

Frequently asked questions

Is link building still worth doing for AI search?

Yes, but be selective. Links still influence which pages get indexed, ranked and retrieved, and the pages that link to you also corroborate how your brand is described. Spend on placements an editor chose for a reader, and drop anything that exists only because it was sold.

Do unlinked brand mentions help AI visibility?

They can. An answer engine can read a mention in context without needing a hyperlink, so a credible page that names you in the right category adds to the picture of who you are. Treat reclaiming links from those mentions as a bonus, and count the mention itself as a result.

How many links do I need to get cited in AI answers?

There's no number, and anyone quoting one is guessing. Citation depends on whether your page is retrievable, whether a passage answers the question cleanly, and whether the source looks trustworthy. Focus on a handful of strong pages and strong placements instead of a link count.

How do I know if my link building is helping AI visibility?

Track a fixed set of buyer questions across the assistants you care about, and log whether you're named and cited over several weeks. Compare that trend with your outreach calendar. It won't prove cause, but a consistent rise after targeted work is reasonable evidence.

What should I cut first?

Start with whatever is cheapest per link and hardest to defend to a reader: bulk guest posts, directory submissions and undisclosed paid placements. Move that budget into original research, PR with a real hook, and fixing the authority you already hold.

Sources

https://www.searchenginejournal.com/link-building-strategies-for-2027-what-to-keep-and-cut/592119/
https://www.searchenginejournal.com/ai-search-is-changing-the-websites-job-not-replacing-it/592134/
https://ahrefs.com/blog/ai-search-roi/
https://www.conductor.com/academy/aeo-geo-benchmarks-report/