seo-evidence
Skill1 starsChanged 31 days agoMITView on GitHub
Grades whether an SEO, AEO or GEO tactic works, using the 182-entry Meta Study ledger and four practitioner layers from anonymised sources with consensus counts, and designs experiments with holdouts, revert triggers and title or snippet tests. Use for "does X work", "is this a myth", "the study says", "did our change work", "should we roll back". Never turns a published gain into a forecast. Not for engine rules (seo-engine-policies).
In this repository7 other files it gives its agents.
- seo-ai-searchSkill
- seo-authoritySkill
- seo-contentSkill
- seo-engine-policiesSkill
- seo-optimizerSkill
- seo-strategySkill
- seo-technicalSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Never forecast a published number onto a client site.
- Very specific: line 15
No rollout without a pre-declared revert rule (applicability, holdout, what stays fixed, metric + window, trigger).
- Distinctive: line 50
Every test declares its revert trigger before launch.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
seo-evidence
Whether a tactic works, how sure we are, and how to test it on this site before anyone calls it a win.
<!-- grammar-excerpt -->Label every tactic with one of the 9 marks: [documented: <URL>, <date>] · [documented-secondary: …] · [unverified: …] · [measured: MARK, n=S, c=C] · [practitioner: anonymised sources · …] · [hypothesis] · [vendor] · [memory] · [unvalidated house heuristic].
Numbers keep their denominator; marks are counts, not verdicts; bundles support direction, not size.
Never forecast a published number onto a client site; only the client's own GSC demand and funnel forecast.
No rollout without a pre-declared revert rule (applicability, holdout, what stays fixed, metric + window, trigger).
Repetition never proves; full grammar: skill seo-optimizer § Evidence marks.
<!-- /grammar-excerpt -->Load rule
Open one reference per load-when row the question touches (a "by H2" row: only the part holding that H2); never a file "to be thorough". A pointer into another seo- skill's file is a citation: quote it, or first load that skill's SKILL.md and open the file through its own load-when row. Engine pages, ledger and practitioner parts and the crawler registry are lookup-only.
Load when
| ID | Question shape | Open |
| EV1 | "does X work", "is it a myth", "is it proven", folklore | by H2: part 1 (the joined table, one row per claim) or part 2 (field errors repeated in popular SEO skills) |
| EV2 | a named ledger entry or study figure, or a Next: pointer from EV1 | the part holding the entry: part 1 (how to read marks, recipe, shortlist), part 2 (catalogue #1–#182), part 3 (AVOID, IT DEPENDS, untested) |
| EV3 | practitioners or forum consensus named in the question | practitioner-index.md, then one of: technical-1, technical-2, agentic, professional, general, digest-1, digest-2 |
| EV4 | test design, holdouts, revert, "did our change work" | the part holding the H2: part 1 (why, recipe), part 2 (designs, measurement, revert triggers) |
| EV5 | titles, snippets, SERP-feature tests | titles-and-serp-tests.md (title ledger #154–#173, guardrails) |
| EV6 | metadata at scale, templated titles, swap test (a single title A/B test is EV5 + EV4) | metadata-at-scale.md |
Not mine
| Question | Skill |
| what an engine forbids, manual actions, appeals | seo-engine-policies |
| implementing schema, canonicals, sitemaps | seo-technical |
| measuring AI citations | seo-ai-search |
| writing the brief or page | seo-content |
Rules
- Quote a ledger entry as mark + n + c + metric + design, never as a bare percentage.
- MAINSTREAM means "published ≥3 times", never "proven". A range crossing zero is "direction unresolved".
[practitioner]claims carry their counts (threads, score, newest month) and stay[practitioner]until a measurement with n and c exists or an engine page says it. Consensus = ≥2 threads, ≥2 authors, ≥10 summed score;contestedalways prints its counter-argument.- A claim repeated across many GitHub skills or blogs is still one claim; star counts are not evidence.
- Two verdicts per URL — position and CTR — never one label ORing them. "Not significant" is not "no effect".
- Every test declares its revert trigger before launch; a human confirms the mechanism before reverting.
- Echo clustering: many sources repeating one upstream origin count as one source, not corroboration; group by origin before counting (skill
seo-content› briefs-and-personas).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.