E-E-A-T is not a score you optimise; it's the set of things your site can prove. For a niche directory, four proofs matter: where every data point comes from, when it was last verified, who runs the site, and what gets a listing added or removed. Ship those four and you're ahead of 90% of directories, most of which are anonymous affiliate shells — which is precisely why Google's systems are tuned to be suspicious of the format.
What E-E-A-T actually is, mechanically
Experience, Expertise, Authoritativeness, Trust come from Google's Search Quality Rater Guidelines — instructions to human raters whose judgments train and validate ranking systems. There is no E-E-A-T ranking factor; there are dozens of signals that correlate with what raters would call trustworthy, learned at scale. The practical consequence: you can't optimise E-E-A-T directly, but you can ship the observable artifacts that trustworthy sites have and spammy ones don't. Google's helpful content guidance is essentially a checklist of those artifacts phrased as questions.
Directories start from a trust deficit. The rater guidelines specifically discuss thin affiliate pages, and a directory is one bad decision away from being one. Everything below is about being legibly not-that.
Verifiable data is a directory's experience signal
A blogger proves experience with photos of the thing they tested. A directory proves it with data that a reader can check in ten seconds. Every app page on this site shows GitHub stars, last-release date, licence, and container-image details pulled via API on a schedule, each with a visible "checked on" date. Any reader can click through to the repository and confirm the numbers match. That reproducibility is the signal — hand-typed specs from 2023 that no longer match reality are the anti-signal, and stale data is the single most common way directories rot.
Three implementation details that matter more than they look:
- Per-field freshness, not per-page. "Stars checked 2026-06-30, description reviewed 2026-02-10" is honest; a page-level "updated yesterday" stamp on unchanged content is the fake freshness pattern that costs you trust with both readers and crawlers.
- Show your nulls. "Licence: could not be determined" beats silently omitting the field. Admitting gaps is a trust signal; papering over them is detectable.
- Cite outward. Linking each fact to its primary source (the repo, the docs, the announcement) is what careful publications do, and directories that never link out look like link hoarders.
Publish a methodology page and link it everywhere
One page, written once, linked from the footer of all 54,000 pages. It should answer, concretely:
- Inclusion criteria. What qualifies a project for listing (e.g. open-source licence, self-hostable, active within 18 months) and what gets one removed.
- Data sources and cadence. Which fields come from which APIs and how often each refreshes.
- Ranking logic. What "best" means on best-of pages — the actual factors, even roughly weighted.
- Money disclosure. Whether anything is sponsored, whether links are affiliate, whether placement is purchasable. If the answer is "no paid placement", say so explicitly; it's rare enough to be a differentiator.
- Corrections. How to report an error and how fast you fix it.
This page does double duty: raters and classifiers see a site that explains itself, and users who land deep in the catalogue can establish in one click why they should believe the page they're on.
Authorship without fake personas
The temptation is to invent an editorial team — "reviewed by our panel of infrastructure experts" over stock headshots. Don't. Fabricated authors are a known spam pattern, AI-generated headshots are trivially recognisable, and a persona with no footprint anywhere else on the web is worse than no byline at all.
The honest version works fine:
| Content type | Attribution that fits |
|---|---|
| Data-driven pages (specs, comparisons) | Site-level attribution + methodology link. No fake byline needed. |
| Editorial verdicts, reviews, blog posts | One real named person with a real history — GitHub, talks, other writing |
| Community-sourced corrections | Credit the reporter, date the fix |
One genuine maintainer with a verifiable identity beats twelve invented experts. If that person has actually run the software — and says so with specifics — that's the E in E-E-A-T, and it's the part a template can't fake.
Trust plumbing: the boring checklist
| Signal | Implementation | Effort |
|---|---|---|
| Contact page | Real email, answered | An hour |
| About page | Who, why, since when | An hour |
| Methodology page | As above | Half a day |
| Corrections policy | One paragraph plus a changelog habit | An hour |
| HTTPS, no intrusive ads | Table stakes | — |
| Outbound citations | Link every fact to source | Template change |
| Field-level "checked on" dates | From your data pipeline | A day |
None of this is clever, which is the point: these signals work because they're costly for spam operations to maintain and cheap for honest sites to ship.
What actually moved for this site
Correlation, not proof, offered as such: the two changes that coincided with this site's recovery in a 2025 core update were per-field provenance (source link plus checked-date on every data point) and removing ~1,200 listings that no longer met the stated inclusion criteria. Enforcing your own methodology page is itself a signal — a directory that visibly culls dead projects reads as maintained, and raters are explicitly told to check whether a site does what it claims.
Bottom line
Ship in this order: methodology page, field-level freshness dates, outbound citations, honest authorship, then the plumbing. Total cost is maybe a week, none of it requires new content, and all of it compounds — every future page inherits the trust infrastructure. What you must not do is easier to state: no invented authors, no fake update stamps, no undisclosed paid placement. E-E-A-T for a directory is mostly the discipline of being auditable, and the sites that lose big in core updates are the ones that couldn't survive an audit.