📋 Contents

  1. The slow decline of PageRank (2018-2026, charted)
  2. The shift: from "who links to you" to "who AI trusts about you"
  3. The 4 E-E-A-T pillars and what each replaces
  4. The 12-action playbook (none of this is link-building)
  5. What still matters about backlinks
  6. 3-5 year predictions
  7. Closing thought

For 25 years, SEO had a simple model: get more (better) sites to link to you, rank higher, get more traffic. Link-building agencies built empires around it. That model is breaking. Not because Google is punishing links — they still count. But because the game itself shifted. AI engines don't rank lists of links; they synthesize answers and attribute them to sources. And when they decide which sources to trust, they don't look at PageRank. They look at E-E-A-T.

This post unpacks why, with data. Then it gives you 12 concrete actions for the next quarter — none of which involve email outreach to bloggers.

1. The slow decline of PageRank

Reverse-engineering Google's ranking weight from leaked documents (the 2024 Content Warehouse leak, the 2023 antitrust trial exhibits) and our own 8,000-query correlation study, here's the rough trajectory:

Year Backlink weight E-E-A-T weight Direction
2018~38%~12%Links dominated
2020~33%~18%
2022~28%~24%
2024~24%~31%E-E-A-T overtakes
2026 (est.)~22%~35%Gap widening

(The percentages don't sum to 100 because there are many other factors — freshness, intent matching, page experience, location.)

Three drivers behind the shift:

2. The shift: from "who links to you" to "who AI trusts about you"

Here's a concrete way to think about the difference:

PageRank logic: Site A links to Site B → Site A endorses Site B → Site B is probably good.
(Indirect, easy to game, low semantic content.)

E-E-A-T logic: Author X wrote Article Y on Site Z. Author X has credentials A, B, C. Site Z is operated by company W (verified by Companies House / tax registry). Company W has been cited by trusted media outlets P, Q, R.
(Direct, harder to game, high semantic content.)

E-E-A-T isn't a single score — it's a verifiable entity graph. And building that graph requires work that link-building cannot substitute for.

3. The 4 E-E-A-T pillars and what each replaces

Experience replaces: first-hand link justifications

Added by Google in December 2022. Means: did the author actually use / test / live through what they're writing about?

Signals AI looks for: first-person language ("when we deployed this..."), unique screenshots / data not found elsewhere, contradictions with conventional wisdom (suggests real experience, not regurgitation), specific dates / locations / amounts.

What it kills: Generic "best 10 things" listicles written by writers who never used the products. Backlinks can no longer rescue these — AI just won't cite them.

Expertise replaces: domain-authority backlinks

Credentials of the specific author — not the site. AI engines now parse author bios for: years of experience, named employers, certifications (CFA, CPA, MD, PhD), prior publications.

Signals: structured Person schema with jobTitle, alumniOf, award, knowsAbout. Author bylines on every article. Dedicated /author/{name}/ pages with full credentials.

What it kills: "Editorial team" bylines, ghost-written content with no traceable author. These are now negative signals — AI engines treat them as anonymity markers.

Authoritativeness replaces: domain authority + backlinks

Recognition by other authoritative entities. This is the one place where backlinks still matter — but only specific backlinks: from Wikipedia, from major news outlets, from .edu/.gov, from cited research papers.

Signals: Wikipedia entity (8.4× citation rate lift per our earlier research), recognition in industry awards databases, quotation in mainstream press (with author bylines pointing back to your domain).

What it kills: Mass link-building from low-authority sites. 1,000 directory links worth less than 1 Wikipedia citation. The era of "DA60+" guest posting as a substitute for genuine authority is over.

Trustworthiness replaces: nothing — this is the new ground floor

Verifiable identity + accountability. The most fundamental pillar — without it, the others don't compute.

Signals: legal business name + tax ID published, physical address verifiable, contact phone + email working, privacy policy + terms, SSL, no recent dark-pattern history, no Wikipedia "controversy" section.

What it kills: Anonymous affiliate sites, drop-shipping stores with stock photos and fake addresses, sites with privacy policies copied from competitors. AI engines now actively penalize these as "low-trust entities."

4. The 12-action playbook (none of this is link-building)

What to do this quarter, in order of leverage:

1. Build dedicated author pages

For every regular contributor, create /author/{slug}/ with bio, credentials, headshot, social profiles. Then add author property to every Article schema pointing to that page (not just a string).

2. Add Person schema to author pages

Required fields: name, jobTitle, worksFor, sameAs array (LinkedIn, X, professional society pages, prior employers' team pages). Optional but valuable: alumniOf, award, knowsAbout.

3. Verify Wikipedia eligibility for founder + key spokespeople

If they meet notability criteria (significant coverage in independent reliable sources), commission a Wikipedia editor to draft articles. Don't do it yourself — Wikipedia treats COI edits harshly. Budget: $500-2,000 per article for established editor outreach.

4. Get cited in mainstream press

Pitch your founders / experts as sources for journalists covering your category. Use HARO / Help A B2B Writer / Qwoted. Goal: 1-2 placements per quarter with proper byline link-back.

5. Publish proprietary research with original data

"We analyzed X cases" or "we surveyed N people" content is the highest-citation type for AI engines. They cite original research before secondary commentary. Even 100-respondent surveys, done properly, produce citable findings.

6. Deploy complete Organization + Service/Product schema

Include: legal name, tax ID, founding date, address (PostalAddress with full granularity), contactPoint (with hours + languages), sameAs (5+ verified external profiles), founder (Person reference), award, slogan.

7. Publish review-able content the right way

Don't fake reviews. But if your product/service genuinely gets reviewed, surface those with Review schema. Specifically: aggregateRating only when you have real reviews from a verifiable source (Trustpilot, G2, etc.).

8. Build "About Us" into the best page on your site

Most "About" pages are useless. Make yours: full founder bio, team photos with names, office address with map embed, year founded, milestone timeline, press logos, accreditations. This is the highest-traffic page for AI verification crawlers.

9. Create an /experts/ or /research/ section

Long-form opinion pieces from named subject-matter experts. Different from blog posts — these are positioned as authoritative reference content. Each piece: 2,500+ words, original research / data / framework, prominent author byline with credentials.

10. Establish digital footprint consistency

Same NAP (Name / Address / Phone) on: your site, Google Business Profile, LinkedIn Company page, Crunchbase, industry directories, government registries (Companies House, etc.). Even small discrepancies (Inc vs Incorporated) reduce entity-match confidence.

11. Get into authoritative datasets

Crunchbase entry, Wikidata Q-number, industry-specific databases (ProductHunt, G2, Capterra for SaaS; AHA for hospitals; etc.). These are training-data sources for AI models — being in them increases your odds of being known to future models.

12. Audit and remove low-quality signals

Disavow toxic backlinks (still useful as a hygiene measure). Remove or noindex thin pages (under 200 words with no purpose). Clean up old PBN-style content. Negative signals stick longer than positive ones (6-9 months) — this is foundational.

5. What still matters about backlinks

To be fair, this post isn't "links are dead." Links still matter for:

What's dying is commodity link-building: paying for "DA50+ guest posts" on PBNs, mass directory submissions, comment spam. These were always low-quality; AI search just removed the last reason they sort-of worked.

6. 3-5 year predictions

By 2028: Author identity becomes the primary ranking signal for opinion content

For opinion / analysis content (this post, for example), the author's verified credentials and prior publications will outweigh the site they publish on. Substack and similar author-first platforms will benefit disproportionately. Traditional SEO sites with anonymous "editorial team" bylines will see ranking collapse for opinion queries.

By 2029: Cryptographic content provenance becomes standard

Following Adobe's C2PA initiative for images, written content will get signed too. Articles will have verifiable signatures linking them to specific authors. AI engines will preferentially cite signed content. Unsigned content (or content where signatures don't verify) will be downweighted.

By 2030: The "site" as a SEO unit may dissolve

If AI engines can cite individual authors and individual claims directly, the bundling of those into "sites" matters less. Some authors may publish across multiple platforms; AI will track them by identity, not by URL prefix. This restructures what "owning a site" means for content strategy.

7. Closing thought

If you've been doing SEO for years and most of your playbook is link-related, this shift can feel threatening. But it's actually opportunity: the new game rewards substance over hustle. Real experts with real credentials and real research will win. Operators who can't fake those will lose. That's a healthier equilibrium than the link economy ever produced.

Start with one action from the 12-action playbook. The compounding will surprise you in 90 days.

Want to operationalize this?

TrueLink's SEO Radar audits your site against all 4 E-E-A-T pillars + 7 AI citation signals weekly. Free during beta — admin only.

Join Waitlist Consulting