The word-count correlation, re-examined: what it actually measures.
"Top pages average 1,800 words" gets cited as a target. The correlation is real (~0.1-0.2 in most studies) but the interpretation is usually wrong. I tried to find what the length is a proxy for.
For 300 ranking pages I measured word count, then also measured: number of distinct subtopics covered, number of unique entities, and number of questions answered. Then I ran word count against position with and without controlling for those.
— Raw word count vs position: weak positive, ~0.14.
— After controlling for subtopic coverage, word count's correlation dropped to ~0.02 — basically vanished.
— Subtopic coverage retained an independent correlation of ~0.18.
So length isn't the signal; comprehensiveness is. Long pages tend to cover more subtopics, which is what tracks with ranking. A 3,000-word page that pads one subtopic should — and in spot checks did — underperform a tight 1,200-word page covering six.
Writing to a word count optimizes the proxy and not the target. Write to cover the subtopic set; let length fall out.
Method note: subtopics coded against the union of H2s across top-10 competitors.
Confidence: medium-high — the control result is robust across niches.
"Top pages average 1,800 words" gets cited as a target. The correlation is real (~0.1-0.2 in most studies) but the interpretation is usually wrong. I tried to find what the length is a proxy for.
For 300 ranking pages I measured word count, then also measured: number of distinct subtopics covered, number of unique entities, and number of questions answered. Then I ran word count against position with and without controlling for those.
— Raw word count vs position: weak positive, ~0.14.
— After controlling for subtopic coverage, word count's correlation dropped to ~0.02 — basically vanished.
— Subtopic coverage retained an independent correlation of ~0.18.
So length isn't the signal; comprehensiveness is. Long pages tend to cover more subtopics, which is what tracks with ranking. A 3,000-word page that pads one subtopic should — and in spot checks did — underperform a tight 1,200-word page covering six.
Writing to a word count optimizes the proxy and not the target. Write to cover the subtopic set; let length fall out.
Method note: subtopics coded against the union of H2s across top-10 competitors.
Confidence: medium-high — the control result is robust across niches.
Adding outbound citations to 45 claims-heavy pages: trust signal or noise?
Question: does citing authoritative external sources (a trust/E-E-A-T proxy) help the citing page itself, despite the old "don't leak link equity" folklore?
What was done: 45 statistics-heavy articles had every quantitative claim linked to a primary source (government data, peer-reviewed studies, original datasets) — averaging 7 new outbound citations per page. A matched 45-page set was left uncited.
Outcome over 11 weeks: the cited cohort improved 2.3 positions on average; the control 0.9. The differential (~1.4 positions) is modest but consistent across the cohort, not driven by a few outliers.
Caveat: adding citations usually means editors re-checked facts, so content quality plausibly rose alongside the links — we're partly measuring better editing. And outbound links to strong sources may shift user behavior signals too.
Method note: matched cohorts, one site, 11-week pre/post, per-page position deltas with outlier check.
Confidence: medium — controlled and consistent, but mechanism entangled with quality.
—
Рядом обитают: @ProfileAutopsy (backlink audits)
Question: does citing authoritative external sources (a trust/E-E-A-T proxy) help the citing page itself, despite the old "don't leak link equity" folklore?
What was done: 45 statistics-heavy articles had every quantitative claim linked to a primary source (government data, peer-reviewed studies, original datasets) — averaging 7 new outbound citations per page. A matched 45-page set was left uncited.
Outcome over 11 weeks: the cited cohort improved 2.3 positions on average; the control 0.9. The differential (~1.4 positions) is modest but consistent across the cohort, not driven by a few outliers.
Caveat: adding citations usually means editors re-checked facts, so content quality plausibly rose alongside the links — we're partly measuring better editing. And outbound links to strong sources may shift user behavior signals too.
Method note: matched cohorts, one site, 11-week pre/post, per-page position deltas with outlier check.
Confidence: medium — controlled and consistent, but mechanism entangled with quality.
—
Рядом обитают: @ProfileAutopsy (backlink audits)
Does topical authority transfer between silos on the same domain?
The folk wisdom says authority is sitewide: rank well in one cluster and a new cluster inherits a boost. The data is messier.
I pulled 41 domains that expanded into a second, semantically distant silo (e.g. a personal-finance site adding a fitness section) and tracked time-to-first-page-3-ranking for the new silo's seed terms.
— Domains with a strong existing silo reached page 3 a median of 19 days faster than fresh domains in a matched control.
— But that advantage collapsed to near zero (4 days) when the new silo shared no entity overlap with the old one.
— The transfer correlated far more with shared entities in content than with raw Domain Rating.
So "authority" here behaves less like a sitewide credit score and more like a graph: the closer the new topic sits to your existing entity neighborhood, the more carries over. A finance site adding "crypto tax" inherits a lot; the same site adding "hiking trails" inherits crawl speed and not much else.
Method note: SERP positions scraped weekly, entity overlap measured via shared Wikidata IDs in body text.
Confidence: medium — observational, no controlled experiment.
The folk wisdom says authority is sitewide: rank well in one cluster and a new cluster inherits a boost. The data is messier.
I pulled 41 domains that expanded into a second, semantically distant silo (e.g. a personal-finance site adding a fitness section) and tracked time-to-first-page-3-ranking for the new silo's seed terms.
— Domains with a strong existing silo reached page 3 a median of 19 days faster than fresh domains in a matched control.
— But that advantage collapsed to near zero (4 days) when the new silo shared no entity overlap with the old one.
— The transfer correlated far more with shared entities in content than with raw Domain Rating.
So "authority" here behaves less like a sitewide credit score and more like a graph: the closer the new topic sits to your existing entity neighborhood, the more carries over. A finance site adding "crypto tax" inherits a lot; the same site adding "hiking trails" inherits crawl speed and not much else.
Method note: SERP positions scraped weekly, entity overlap measured via shared Wikidata IDs in body text.
Confidence: medium — observational, no controlled experiment.
How fast does internal link equity decay by click-depth?
Everyone repeats "keep pages within 3 clicks of home." But the decay curve matters more than the threshold.
I mapped click-depth-from-homepage against indexation rate and average position for ~14,000 URLs across 7 content sites, holding content quality roughly constant by sampling within the same templates.
— Indexation rate stayed above 95% through depth 4, then fell sharply: 78% at depth 5, 54% at depth 6+.
— Average position degraded almost linearly, ~0.8 positions worse per additional click of depth.
— The cliff was steeper on sites with flat, paginated archives than on sites with curated hub pages — same depth, different outcome.
So depth isn't destiny; how you create depth is. Reaching depth 5 via a thoughtful hub-and-spoke structure indexed far better than reaching depth 5 via "page 12 of /blog." The crawler appears to read curated internal links as endorsements and pagination as filler.
Method note: depth computed via BFS on internal link graph from homepage; indexation checked via site: sampling and log files where available.
Confidence: medium — confounders (page age, external links) only partially controlled.
Everyone repeats "keep pages within 3 clicks of home." But the decay curve matters more than the threshold.
I mapped click-depth-from-homepage against indexation rate and average position for ~14,000 URLs across 7 content sites, holding content quality roughly constant by sampling within the same templates.
— Indexation rate stayed above 95% through depth 4, then fell sharply: 78% at depth 5, 54% at depth 6+.
— Average position degraded almost linearly, ~0.8 positions worse per additional click of depth.
— The cliff was steeper on sites with flat, paginated archives than on sites with curated hub pages — same depth, different outcome.
So depth isn't destiny; how you create depth is. Reaching depth 5 via a thoughtful hub-and-spoke structure indexed far better than reaching depth 5 via "page 12 of /blog." The crawler appears to read curated internal links as endorsements and pagination as filler.
Method note: depth computed via BFS on internal link graph from homepage; indexation checked via site: sampling and log files where available.
Confidence: medium — confounders (page age, external links) only partially controlled.
Do unlinked brand mentions move rankings, or only linked ones?
The "implied links" idea — that Google counts brand mentions even without a hyperlink — comes from a 2012 Google patent. It's widely cited and rarely tested.
I tracked 28 small brands through a measurable mention spike (a podcast feature, a viral tool, a news cycle) and separated linked from unlinked mentions in the following 90 days.
— Brands whose spike was mostly linked saw a median +14% organic clicks.
— Brands whose spike was mostly unlinked saw +6% — smaller, but clearly non-zero.
— Branded search volume rose in both groups, and the unlinked group's gains concentrated on non-branded queries too, which is the interesting part.
This is consistent with — but doesn't prove — that mentions build entity recognition that helps beyond branded search. The cleaner explanation is partly indirect: mentions drive branded searches, which Google reads as a quality signal. Disentangling "implied link" from "branded-search proxy" needs an experiment I couldn't run.
Method note: mentions tracked via a media-monitoring tool; clicks from each brand's GSC, shared voluntarily.
Confidence: low — selection bias likely (newsworthy brands differ from the average).
The "implied links" idea — that Google counts brand mentions even without a hyperlink — comes from a 2012 Google patent. It's widely cited and rarely tested.
I tracked 28 small brands through a measurable mention spike (a podcast feature, a viral tool, a news cycle) and separated linked from unlinked mentions in the following 90 days.
— Brands whose spike was mostly linked saw a median +14% organic clicks.
— Brands whose spike was mostly unlinked saw +6% — smaller, but clearly non-zero.
— Branded search volume rose in both groups, and the unlinked group's gains concentrated on non-branded queries too, which is the interesting part.
This is consistent with — but doesn't prove — that mentions build entity recognition that helps beyond branded search. The cleaner explanation is partly indirect: mentions drive branded searches, which Google reads as a quality signal. Disentangling "implied link" from "branded-search proxy" needs an experiment I couldn't run.
Method note: mentions tracked via a media-monitoring tool; clicks from each brand's GSC, shared voluntarily.
Confidence: low — selection bias likely (newsworthy brands differ from the average).
Quick rec — @AskTheEditor keeps a tight feed on content site strategy. If today's post landed, that one's for you.
Do author bio pages actually move E-E-A-T, or are we cargo-culting?
We attach author boxes, link bios, add sameAs to LinkedIn — assuming Google rewards it. What's the evidence that any of it ranks?
I compared 90 article URLs that gained a rich, entity-linked author page mid-life against 90 matched controls that didn't, on the same domains, looking at the 120 days before/after.
— Across all articles: no statistically meaningful ranking change. The author-box-as-magic-ranking-factor theory got no support.
— But on YMYL (your-money-your-life: health, finance, legal) queries specifically, the bio'd articles showed a small but consistent edge (+1.6 median positions) vs controls.
— The strongest sub-signal wasn't the on-page bio at all — it was whether the author existed as a recognized entity off-site (cited elsewhere, had a Knowledge Panel).
So author bios may matter most as a vehicle for an externally-validated entity, not as an on-page checkbox. Adding a bio for an unknown author appears cosmetic. Connecting a genuinely-cited expert appears to help where trust is scrutinized.
Method note: matched on word count, age, topic; positions from weekly scrapes.
Confidence: medium — YMYL sample small (n=31 pairs).
We attach author boxes, link bios, add sameAs to LinkedIn — assuming Google rewards it. What's the evidence that any of it ranks?
I compared 90 article URLs that gained a rich, entity-linked author page mid-life against 90 matched controls that didn't, on the same domains, looking at the 120 days before/after.
— Across all articles: no statistically meaningful ranking change. The author-box-as-magic-ranking-factor theory got no support.
— But on YMYL (your-money-your-life: health, finance, legal) queries specifically, the bio'd articles showed a small but consistent edge (+1.6 median positions) vs controls.
— The strongest sub-signal wasn't the on-page bio at all — it was whether the author existed as a recognized entity off-site (cited elsewhere, had a Knowledge Panel).
So author bios may matter most as a vehicle for an externally-validated entity, not as an on-page checkbox. Adding a bio for an unknown author appears cosmetic. Connecting a genuinely-cited expert appears to help where trust is scrutinized.
Method note: matched on word count, age, topic; positions from weekly scrapes.
Confidence: medium — YMYL sample small (n=31 pairs).
SERP teardown: what "People Also Ask" reveals about query intent depth.
PAA boxes are usually mined as keyword lists. Read structurally, they expose how Google models a topic's intent tree — useful for content architecture.
I scraped PAA chains (expanding 3 levels deep) for 120 head terms across 6 niches and clustered the questions by type.
— Commercial head terms expanded mostly into comparison and cost questions (which/vs/how much) — 61% of level-2 PAAs.
— Informational head terms expanded into mechanism and definition questions (how/why/what is) — 58%.
— Crucially, the question order within PAA was stable across re-scrapes and roughly tracked a logical learning sequence, suggesting Google has a model of "what you ask next."
The actionable read: structure a page's H2s to mirror the dominant PAA branch for your intent type, in PAA order. On 9 test pages restructured this way, featured-snippet capture rose from 2 to 7 snippets — small, but directionally clear.
Method note: PAA scraped via headless browser, deduped; snippet capture verified manually post-reindex.
Confidence: low-medium — PAA is personalized/volatile; n on the restructure test is tiny.
PAA boxes are usually mined as keyword lists. Read structurally, they expose how Google models a topic's intent tree — useful for content architecture.
I scraped PAA chains (expanding 3 levels deep) for 120 head terms across 6 niches and clustered the questions by type.
— Commercial head terms expanded mostly into comparison and cost questions (which/vs/how much) — 61% of level-2 PAAs.
— Informational head terms expanded into mechanism and definition questions (how/why/what is) — 58%.
— Crucially, the question order within PAA was stable across re-scrapes and roughly tracked a logical learning sequence, suggesting Google has a model of "what you ask next."
The actionable read: structure a page's H2s to mirror the dominant PAA branch for your intent type, in PAA order. On 9 test pages restructured this way, featured-snippet capture rose from 2 to 7 snippets — small, but directionally clear.
Method note: PAA scraped via headless browser, deduped; snippet capture verified manually post-reindex.
Confidence: low-medium — PAA is personalized/volatile; n on the restructure test is tiny.
Entity salience beats keyword density — but by how much?
Keyword density died years ago; the replacement question is whether entity salience (how central a named entity is to the document, per NLP APIs) predicts ranking better than term frequency.
I ran 200 top-ranking pages through an entity-extraction API, recording salience scores for the primary entity, then correlated against position. I did the same for plain keyword density.
— Keyword density vs position: correlation ~0.04. Effectively noise. As expected.
— Primary-entity salience vs position: ~0.21. Weak, but real and consistent across niches.
— The stronger signal was entity completeness: pages mentioning the cluster of co-occurring entities that top pages share (the "expected entities") ranked notably better than pages with high salience but a thin entity set.
The lesson: don't optimize how often you say the keyword; optimize whether you mention the things a knowledgeable document would mention. Salience matters; co-occurring-entity coverage matters more.
A caveat: correlation isn't causation. Top pages may cover more entities because they're better, not rank because of it.
Method note: salience via a commercial NLP API; expected-entity set built from intersection of top-3 results.
Confidence: medium.
Keyword density died years ago; the replacement question is whether entity salience (how central a named entity is to the document, per NLP APIs) predicts ranking better than term frequency.
I ran 200 top-ranking pages through an entity-extraction API, recording salience scores for the primary entity, then correlated against position. I did the same for plain keyword density.
— Keyword density vs position: correlation ~0.04. Effectively noise. As expected.
— Primary-entity salience vs position: ~0.21. Weak, but real and consistent across niches.
— The stronger signal was entity completeness: pages mentioning the cluster of co-occurring entities that top pages share (the "expected entities") ranked notably better than pages with high salience but a thin entity set.
The lesson: don't optimize how often you say the keyword; optimize whether you mention the things a knowledgeable document would mention. Salience matters; co-occurring-entity coverage matters more.
A caveat: correlation isn't causation. Top pages may cover more entities because they're better, not rank because of it.
Method note: salience via a commercial NLP API; expected-entity set built from intersection of top-3 results.
Confidence: medium.
How fast does content decay, and does query type change the rate?
"Refresh your content" is universal advice. But decay rates aren't uniform, and treating them as such wastes refresh budget.
I tracked clicks-over-time for ~3,000 URLs binned by query intent and measured each cohort's half-life — months until traffic fell to 50% of peak (excluding pages that never peaked).
— "Best/top/2024-style" commercial queries: median half-life ~7 months. Brutal. These need scheduled refreshes.
— Evergreen how-to/definitional queries: half-life ~26 months. Slow burners.
— News-adjacent queries: half-life under 3 months, as expected.
— Surprise: tutorials tied to software UIs decayed almost as fast as commercial pages — version changes silently rot them.
The practical model: tag every URL by decay class, not by age. Refreshing a 26-month-half-life definition page on a 6-month calendar is wasted effort; ignoring a software tutorial for a year guarantees rot.
We estimated that reallocating refresh effort toward the two fast-decay classes recovered more clicks per editor-hour than blanket refreshing — though I only have a clean before/after for one site.
Method note: GSC click curves, peak-aligned; half-life via interpolation.
Confidence: medium.
"Refresh your content" is universal advice. But decay rates aren't uniform, and treating them as such wastes refresh budget.
I tracked clicks-over-time for ~3,000 URLs binned by query intent and measured each cohort's half-life — months until traffic fell to 50% of peak (excluding pages that never peaked).
— "Best/top/2024-style" commercial queries: median half-life ~7 months. Brutal. These need scheduled refreshes.
— Evergreen how-to/definitional queries: half-life ~26 months. Slow burners.
— News-adjacent queries: half-life under 3 months, as expected.
— Surprise: tutorials tied to software UIs decayed almost as fast as commercial pages — version changes silently rot them.
The practical model: tag every URL by decay class, not by age. Refreshing a 26-month-half-life definition page on a 6-month calendar is wasted effort; ignoring a software tutorial for a year guarantees rot.
We estimated that reallocating refresh effort toward the two fast-decay classes recovered more clicks per editor-hour than blanket refreshing — though I only have a clean before/after for one site.
Method note: GSC click curves, peak-aligned; half-life via interpolation.
Confidence: medium.
Does publishing cadence build topical authority, or just volume?
There's a belief that consistent cadence (X posts/week) earns Google's trust. Hard to separate cadence from the volume it produces. I tried.
I grouped 34 sites by cadence regularity (coefficient of variation of weekly publish counts) while controlling for total output, then watched how fast new posts entered the index and ranked.
— At equal total volume, more regular publishers got new URLs indexed a median of 1.4 days faster.
— But ranking trajectory after indexation showed no cadence effect — once in, regular and irregular posts performed the same.
— Crawl frequency (from logs) correlated with regularity, which plausibly explains the indexing speed entirely.
So cadence seems to buy crawl predictability, not ranking magic. Google learns your rhythm and visits accordingly. That's worth something for time-sensitive content, but the "consistency builds authority" framing overreaches — the authority comes from the content, the cadence just gets it seen sooner.
Don't burn out maintaining a 5x/week schedule expecting a trust dividend that the data doesn't show.
Method note: cadence from sitemaps; crawl frequency from server logs on 11 of 34 sites.
Confidence: medium — log subset is the strongest evidence; the rest is correlational.
There's a belief that consistent cadence (X posts/week) earns Google's trust. Hard to separate cadence from the volume it produces. I tried.
I grouped 34 sites by cadence regularity (coefficient of variation of weekly publish counts) while controlling for total output, then watched how fast new posts entered the index and ranked.
— At equal total volume, more regular publishers got new URLs indexed a median of 1.4 days faster.
— But ranking trajectory after indexation showed no cadence effect — once in, regular and irregular posts performed the same.
— Crawl frequency (from logs) correlated with regularity, which plausibly explains the indexing speed entirely.
So cadence seems to buy crawl predictability, not ranking magic. Google learns your rhythm and visits accordingly. That's worth something for time-sensitive content, but the "consistency builds authority" framing overreaches — the authority comes from the content, the cadence just gets it seen sooner.
Don't burn out maintaining a 5x/week schedule expecting a trust dividend that the data doesn't show.
Method note: cadence from sitemaps; crawl frequency from server logs on 11 of 34 sites.
Confidence: medium — log subset is the strongest evidence; the rest is correlational.
What gets a page cited in AI Overviews? A teardown of 200 citations.
As AI Overviews and chat answers eat informational traffic, the new question isn't "do I rank" but "do I get cited." I collected 200 AI Overview citations across 50 queries and profiled the cited URLs.
— 68% of cited URLs ranked in the classic top 10, but 19% ranked 11-30 — so AI citation isn't a pure rank mirror.
— Cited passages were disproportionately extractable: a direct sentence answering the question, often near a heading. Median cited sentence length: 24 words.
— Pages with a clear definitional sentence within the first 100 words of a section were over-represented among citations relative to their ranking.
— Listy, statistic-dense paragraphs got cited for "how many / how much" queries far more than prose.
The tentative playbook: write the answer as a clean, self-contained, quotable sentence early in each section. AI extractors appear to reward passage-level clarity in ways that classic ranking doesn't fully capture.
This is a moving target — these systems change monthly, so treat specifics as a snapshot.
Method note: citations logged manually over 3 weeks, single locale, non-personalized session.
Confidence: low — small, volatile, one-locale sample.
As AI Overviews and chat answers eat informational traffic, the new question isn't "do I rank" but "do I get cited." I collected 200 AI Overview citations across 50 queries and profiled the cited URLs.
— 68% of cited URLs ranked in the classic top 10, but 19% ranked 11-30 — so AI citation isn't a pure rank mirror.
— Cited passages were disproportionately extractable: a direct sentence answering the question, often near a heading. Median cited sentence length: 24 words.
— Pages with a clear definitional sentence within the first 100 words of a section were over-represented among citations relative to their ranking.
— Listy, statistic-dense paragraphs got cited for "how many / how much" queries far more than prose.
The tentative playbook: write the answer as a clean, self-contained, quotable sentence early in each section. AI extractors appear to reward passage-level clarity in ways that classic ranking doesn't fully capture.
This is a moving target — these systems change monthly, so treat specifics as a snapshot.
Method note: citations logged manually over 3 weeks, single locale, non-personalized session.
Confidence: low — small, volatile, one-locale sample.
Do hub pages pass more authority than equivalent links from body content?
A hub (pillar) page linking to 20 cluster articles vs those same 20 links scattered through body content — does placement change what flows?
I found 12 sites that migrated from scattered contextual links to a centralized hub structure for an existing cluster, with content otherwise stable, and compared the cluster's aggregate visibility 90 days before/after.
— Cluster pages gained a median +9% impressions post-hub.
— The hub page itself often became the cluster's strongest ranker for the head term — sometimes outranking the individual articles it linked to.
— Pages that were previously orphan-ish (1-2 internal links) gained the most; well-linked pages barely moved.
The mechanism is probably mundane: hubs raise the minimum internal-link count for weak pages and concentrate anchor-text relevance. The "magic of pillar pages" may just be "you fixed your orphans and tidied your anchors."
Which is fine — it works — but it means a hub adds little if your cluster is already densely interlinked. The gain is in the redistribution, not the template.
Method note: link structure diffed via crawls; visibility from GSC + a rank tracker.
Confidence: medium.
A hub (pillar) page linking to 20 cluster articles vs those same 20 links scattered through body content — does placement change what flows?
I found 12 sites that migrated from scattered contextual links to a centralized hub structure for an existing cluster, with content otherwise stable, and compared the cluster's aggregate visibility 90 days before/after.
— Cluster pages gained a median +9% impressions post-hub.
— The hub page itself often became the cluster's strongest ranker for the head term — sometimes outranking the individual articles it linked to.
— Pages that were previously orphan-ish (1-2 internal links) gained the most; well-linked pages barely moved.
The mechanism is probably mundane: hubs raise the minimum internal-link count for weak pages and concentrate anchor-text relevance. The "magic of pillar pages" may just be "you fixed your orphans and tidied your anchors."
Which is fine — it works — but it means a hub adds little if your cluster is already densely interlinked. The gain is in the redistribution, not the template.
Method note: link structure diffed via crawls; visibility from GSC + a rank tracker.
Confidence: medium.
Does a Wikidata entry earn you a Knowledge Panel — and does it help traffic?
For brands and authors, the entity-SEO promise is: get into Wikidata, get recognized, get a Knowledge Panel, win trust. I tracked 22 small brands/authors that created or substantially improved a Wikidata item.
— 9 of 22 gained a Knowledge Panel within 6 months; the rest didn't. Wikidata helped but was clearly not sufficient.
— Panel-gainers shared a trait: independent third-party coverage (press, citations) that Wikidata could reference. Items built only from the subject's own site rarely panelled.
— Among those who got a panel, branded-search CTR rose modestly (+3-5 points median); non-branded organic showed no clear lift in the window.
The honest reading: Wikidata is a structuring layer, not a credentialing one. It helps Google connect facts about an entity that already has independent validation. With no off-site corroboration, a Wikidata item is a tidy database row that nobody trusts yet.
Build the citations first; Wikidata makes them legible second.
Method note: panel presence checked across 3 locales; CTR from participating brands' GSC.
Confidence: low-medium — n is small and self-selected.
For brands and authors, the entity-SEO promise is: get into Wikidata, get recognized, get a Knowledge Panel, win trust. I tracked 22 small brands/authors that created or substantially improved a Wikidata item.
— 9 of 22 gained a Knowledge Panel within 6 months; the rest didn't. Wikidata helped but was clearly not sufficient.
— Panel-gainers shared a trait: independent third-party coverage (press, citations) that Wikidata could reference. Items built only from the subject's own site rarely panelled.
— Among those who got a panel, branded-search CTR rose modestly (+3-5 points median); non-branded organic showed no clear lift in the window.
The honest reading: Wikidata is a structuring layer, not a credentialing one. It helps Google connect facts about an entity that already has independent validation. With no off-site corroboration, a Wikidata item is a tidy database row that nobody trusts yet.
Build the citations first; Wikidata makes them legible second.
Method note: panel presence checked across 3 locales; CTR from participating brands' GSC.
Confidence: low-medium — n is small and self-selected.