E-E-A-T

1. What E-E-A-T is

E-E-A-T is the property that decides whether the source behind an already-retrieved passage is trustworthy enough to be used. It is not about being found, and not about a passage’s shape — it is about who is behind it.

Definition (GEO Wiki working definition): E-E-A-T is the quality/trust property that decides whether an already-retrieved source is worth grounding an answer on — independent of whether any one of its passages is structurally liftable.

One honesty line up front, and it is load-bearing: E-E-A-T is a term from Google’s Search Quality Rater Guidelines, not an algorithmic score. The fourth letter — Experience — was added in December 2022 (see E-A-T gets an extra E). It describes what human raters look for; it is not a ranking signal and no engine computes an “E-E-A-T number”. Hold that — §5 makes the mechanism honest.

In the taxonomy this is the content-quality signal — the trust/authority half of grounding. Its sibling is Citability, the structure half. The two are orthogonal levers on the same step; §3 makes the boundary explicit. The one-sentence version: trust decides if a source is allowed to be used; structure decides if a passage can be lifted.

2. Trusted ≠ used — why E-E-A-T is its own gate

The claim that justifies a standalone entry: a page can be crawled, retrieved into the candidate set, and structurally perfectly liftable — and still never used, because the source is not trusted. This is the exact mirror of Citability §2’s “retrieved ≠ grounded”: here it is “liftable ≠ trusted”.

  candidate passage set
        │
        ▼
  ┌─────────────────────────────┐
  │  TRUST / E-E-A-T GATE        │
  │  source worthy?             │  ── no ──►  filtered out
  │  author real / corroborated?│             (liftable,
  │  claims verifiable?         │              never selected)
  └─────────────────────────────┘
        │ yes
        ▼
  grounded subset ──► synthesis ──► (maybe) attribution

This is the same gate citability draws — viewed from its other axis. A passage can pass the citability check (self-contained, answer-shaped) and be dropped here because nothing about the source is trustworthy.

E-E-A-T and Citability sit on the same step — grounding/selection, Answer Loop §3.3 “the choke point” — as two independent gates.

3. E-E-A-T vs Citability — the two orthogonal grounding levers

The single highest-value disambiguation on this page, and the exact mirror of Citability §3. Both gate grounding; they are independent. Most “I did everything and still wasn’t used” confusion comes from collapsing these two into one.

The load-bearing line, stated as the reciprocal of citability’s: a perfectly chunked page with no authority loses on trust; a trusted source written as a wall of text still loses grounding. You need both — they do not substitute.

This entry formally owns what citability hands off: author credentials, first-hand experience, and citation density as a quality signal belong here and are not re-taught on the citability page. Structure-as-shape stays there; trust-of-source is this entry’s.

4. The four signals — what each looks like to an AI engine

This is the trust-half taxonomy (mirror of Citability §4’s structural seven). Google states the hierarchy plainly: the four “help determine which content demonstrates… E-E-A-T. Of these aspects, trust is most important. The others contribute to trust” (see Creating Helpful Content).

Experience is the AI-era differentiator. The fourth “E”, added Dec 2022, evaluates “whether content demonstrates that it was produced with some degree of experience, such as with actual use of a product, having actually visited a place or communicating what a person experienced” (Google). It is the hardest signal to fake at scale and the most natural counter to mass-generated content — the bridge to §7 and AI Content Detection.

4.1 Experience

First-hand contact the text could not have been written without.

• ✓ “On the 1,000-URL crawl we ran, 38% blocked GPTBot — here is the breakdown.”
• ✗ “Studies show many sites block AI crawlers.”

4.2 Expertise

A real, resolvable author with domain command — not a byline.

• ✓ Named author, linked bio, consistent identity across platforms (sameAs).
• ✗ “By Admin” or no author, generic phrasing any page could carry.

4.3 Authoritativeness

Recognition that comes from others, not from your own page.

• ✓ Cited by sources the model already trusts; present in the knowledge graph.
• ✗ “The leading authority on…” with nothing external to corroborate it.

4.4 Trustworthiness

The apex Google names — the other three feed it.

• ✓ Claims sourced, dates current, ownership transparent, internally consistent.
• ✗ Unsourced numbers, stale “as of 2021”, contradicted elsewhere on the site.

5. How AI engines actually consume E-E-A-T — the honest mechanism

This is the entry’s honesty section, the trust-half counterpart to Citability §5. The core honesty: there is no “E-E-A-T score”. AI engines do not compute E-E-A-T. They consume a set of trust/authority proxies at three places — retrieval, grounding, and the model’s prior — and E-E-A-T is just the name for the family those proxies fall in.

The seam, stated explicitly: this entry owns the quality/trust framing of these signals — why each is a trust proxy. It routes the entity-graph mechanism — how an identity is actually resolved across platforms — to Entity Recognition, Knowledge Graph Presence, and Brand Mentions. It does not re-teach them here. (This mirrors how citability routes crawler and language mechanics out rather than absorbing them.)

6. What the evidence says — and what it does not

The empirical anchor, carried with the same honesty discipline as the source paper. Aggarwal et al. found that content-substance rewrites — cite sources, add statistics, add quotations — measurably raised answer visibility, while keyword stuffing did not. The trust half is why the ceiling on that lift is bounded: fabricated substance fails the trust filter it imitates.

The position, stated plainly as the reciprocal of citability’s “take the direction, discard the number”: E-E-A-T is earned, not annotated. Genuine first-hand experience and corroborated authority move it; manufactured signals trip the filter. The single-actor lift is an upper bound, not the equilibrium once competitors optimize the same engine (C-SEO Bench, Puerto et al., NeurIPS ‘25 D&B). For the full critique, see the paper entry.

One adjacent line, routed not expanded: whether an engine credits the source it grounded on is a separate, verifiability-and-attribution problem (Liu et al.) — being trusted enough to be used is not the same as being cited.

7. Anti-patterns — fake E-E-A-T and why it backfires

E-E-A-T is the entry most likely to be faked. Each anti-pattern below looks like the signal it imitates and fails because it trips a trust or AI-spam filter — the trust-half counterpart to Citability §6.

The load-bearing line, the reciprocal of citability’s “necessary, not sufficient”: E-E-A-T is necessary, not sufficient. A trusted source whose passages cannot be lifted still loses grounding — that gap is Citability’s. And faked authority is itself a penalized signal: over-claimed, low-substance patterning is covered by AI Content Detection. Google’s own position is that there are no special tricks beyond helpful, original, people-first content (see Creating Helpful Content).

8. E-E-A-T across SEO and GEO — invariant baseline vs what changes

Real E-E-A-T is a shared baseline, not a GEO-specific lever. SEO vs GEO commits this in three places: it is identical in both, it is on the “never drop” list, and dropping it degrades blue links and AI answers at once. This entry restates that contract rather than re-deriving it.

What is invariant: the four signals themselves. What changes is how they are consumed — from a human rater’s quality heuristic to a trust filter applied at retrieval and grounding plus the model’s prior. The signal is the same; the consumer moved.

Two routed lines, not expanded here: trust signals are not language-invariant — corroboration pools and authority cues differ across languages, which is Multilingual GEO’s. And the trust-readability of non-text assets (image provenance, video author signals) is Multimodal Signals’.

9. Why this matters for GEO + how to act

Grounding is the choke point Answer Loop §3.3 calls the highest-leverage step — and E-E-A-T is its trust lever, as Citability is its structure lever. This entry is the concept; the doing is the playbook.

For the term itself and its neighbors, see the GEO glossary.

References

Official (Google):

• Google — General Guidelines (Search Quality Rater Guidelines) (rev. 2025-09-11) — the document that defines E-E-A-T
• Google Search Central — Creating Helpful, Reliable, People-First Content · E-A-T gets an extra E for Experience (2022-12-15)
• Google Search Central — AI features and your website · Succeeding in AI search

Academic:

• Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K. & Deshpande, A. (2024). GEO: Generative Engine Optimization. KDD ‘24. arXiv:2311.09735 · ACM DL · paper summary
• Puerto, H., Gubri, M., Green, C., Oh, S. J. & Yun, S. (2025). C-SEO Bench: Does Conversational SEO Work? NeurIPS ‘25 Datasets & Benchmarks. arXiv:2506.11097
• Liu, N. F., Zhang, T. & Liang, P. (2023). Evaluating Verifiability in Generative Search Engines. Findings of EMNLP 2023. arXiv:2304.09848