SEO Contextual Density Index: How We Rank Sites
The web is flooded with “Top 10 Best…” articles. The reason is simple: these list articles, listicles, rank fast and hard on Google, and the popular AIs lean on them heavily as sources. Google and ChatGPT seem to have reduced their reach into the large language models (LLMs), but their presence is still felt. ProStar SEO has used this technique for years, successfully. Still, I hate an approach where you have to proclaim yourself number one at everything to be recognized by the LLMs.
List articles have their place. If I am shopping for a new lawn mower and know nothing about them, a blog that walks me through the ten best models lets me choose with judgment.
So I decided to change our approach. We are launching the SEO Contextual Density Index, which ranks websites on the measurable factors we consider most important for a site’s search performance. These factors are mostly quantitative and rest on tens of thousands of tests and observations. Their sources are the content of the site’s page, the Semrush database and, for the qualitative side, the reviews on its Google Business Profile.
How the index works
The index answers one question: among the sites competing for a keyword, which ones show the signals that good SEO produces? To answer it, it asks nobody’s opinion. It takes the keyword, keeps the sites that rank a real service page for it, downloads each page, measures it with the same script, adds three Semrush readings and the Google Business Profile, then adds everything up with a formula published further down. The result is a score out of 100 per site, and a table where every column can be read, sorted and verified.
Six signals make up the score. Two come from the text of the page: semantic coverage and contextual density. Three come from Semrush: page-1 keywords, commercial-intent keywords and visibility in Google’s AI answers. The last comes from the Google Business Profile: rating and review count. ProStar SEO is measured with the same instrument as everyone else and appears in every table at its real rank, even when that rank is poor.
The rest of this page explains each piece, starting with the most important one.
The SEO Contextual Density Index: the foundation
Google counts, it doesn’t read
Picture a library. The title on the spine is your title tag, the one that shows in search results. The title on the cover is your H1. The table of contents is your H2s. A reader who opens the book reads the chapters; Google does not read. It counts. It counts the words that belong to the subject, the variety of those words, where they appear, and it compares all of that with the other books on the same shelf.
Google’s spokespeople have repeated for years that content quality is the first ranking factor, without ever saying how an algorithm measures quality. Our answer is two words: contextual density. Google grades a page the way academia grades a scientific paper: it rewards the one that delivers the most relevant information with the least fluff, and that makes a complex problem clear. Whether the page sells a product, describes a service or explains a topic, quality takes the same shape, a rich and exact vocabulary focused on the visitor’s intent. It is measurable, and it is what we measure: the ratio of insight to words.
What the data says
This conviction does not come from a hunch. It comes from years of testing Google’s algorithm with Cora SEO software, which analyzes more than 5,000 factors across the top 100 pages of every query studied. Every search run by the tool’s hundreds of users adds its data points; the correlations between factors and positions are aggregated and ranked by strength. By convention they are negative: a factor whose value rises as the position gets closer to first place shows a negative coefficient. The further from zero, the stronger the link.
The figures below come from one Cora report on a real French-language market, sorted by how strongly each factor tracks position. Read the ordering rather than the exact numbers: a single query is a sample, not a census, and the ordering is the part that has held in every reading we have run since 2025.
Correlation data
Top on-page factors, correlated with ranking position
A bar further to the right means a closer link with the top of the results. Every factor here clears the 95% significance threshold for a 100-result sample.
- Correlation with position
- Significance threshold (0.21)
Source: one Cora SEO 7.2 report on the query “courtier immobilier québec” (Google Canada, French, 100 results), September 1, 2026, plus ProStar SEO in-house analysis. Over 100 results the 95% significance threshold is 0.21, and all sixteen factors reach or exceed it. One market and one query, so read the ordering rather than the exact figures.
Three readings stand out. Vocabulary variety dominates: the number of factors shared with the top results, distinct entities, unique LSI terms and unique keyword variations sit at the top of both columns. Placement comes next: the same variations weigh everywhere they appear, in the H1 to H6 headings, paragraphs, lists, links and sentences, which means a rich vocabulary has to run through the whole page and not one block. And repetition weighs less than variety: on the August 2026 query, the exact-match factors (exact keyword density, exact matches in the HTML, headings or links) sit between −0.13 and −0.19, roughly half the variety factors; the frequency of the subject’s terms, all terms combined, reaches −0.21. Word count (−0.20) reads as a by-product, covering a subject takes words, but length is the consequence, not the cause. As for backlink count, it hovered around 4th place in the aggregate two years ago; today it sits 73rd there (−0.13), and the August 2026 query gives it the same value. Links still matter, but through their relevance and the contextual density of the page carrying them, which a correlation tool measures poorly.
Two limits apply to this data and we would rather state them. Correlation is not causation: pages rich in vocabulary are naturally longer, and it is the richness, not the length, that explains the word-count score. And the sample has its biases: .com domains look better ranked because there are far more of them, not because Google prefers them. For that reason we exclude domain extensions, social signals, which track brand size, and word count, which tracks richness. Contextual density is precisely the answer to the first limit: it relates relevant vocabulary to length, so as to reward richness rather than padding. The metric itself is not a Cora factor; it is validated by our before-and-after tests on client pages, where it improves rankings roughly nine times out of ten when the site’s technical base is sound.
What Google let slip
The May 2024 leak of Google’s API documentation showed metrics named siteFocusScore and siteRadius: Google builds a vector representation of every page and every site, then measures how tightly the content concentrates around a topic and how far each page strays from the site’s core theme. Under another name, that is what contextual density measures. The same documentation describes an entity salience score: Google grades not only which entities appear but how important they are to the text, stronger in a heading or as the subject of a sentence than in a parenthesis. Two consequences for a site: contextual density plays page by page, but also site-wide, where off-topic content dilutes the authority of the whole; and the choice of terms matters as much as their number, a central entity placed at the head of a sentence being worth more than a tangential mention buried in a list.
Contextual density (CD): the original formula
Three families of words count. Keyword variations: the core keyword, its combinations, synonyms, singulars and plurals. Entities: the recognized brands, places, people, tools and concepts that accompany the subject. LSI terms: the action, description and technical words that surround it. We add up all their occurrences and divide by the number of useful words on the page, that is, all words minus stop words (articles, prepositions, pronouns, auxiliaries).
CD = (Σ(KV) + Σ(RE) + Σ(LSI)) / (TW − SW) × 100
Σ(KV): total occurrences of keyword variations, repetitions included. Σ(RE): total occurrences of relevant entities. Σ(LSI): total occurrences of relevant LSI terms. TW: total word count of the page. SW: stop-word count.
The result is a percentage: the share of the useful text that actually carries the subject. A page of 1,000 useful words where the subject’s terms appear 90 times has a CD of 9%.
Contextual density is not keyword density. Keyword density counts the repetition of a single phrase, and the Cora data gives it roughly half the weight of vocabulary variety. Contextual density counts the presence of an entire semantic field, whose components form the family of factors most correlated with rankings.
Unique contextual density (UCD): the added formula
CD has a known weakness: a term used five times counts five. A page can therefore reach a high density by repeating a few words. To see what CD blends together, the index also publishes unique contextual density, which counts each term once. It is added to CD; it does not replace it.
UCD = (U(KV) + U(RE) + U(LSI)) / (TW − SW) × 100
U(KV): number of distinct keyword variations present in the text. U(RE): number of distinct entities present. U(LSI): number of distinct LSI terms present. TW and SW: identical to the CD formula.
Only the numerator changes. On the same page of 1,000 useful words, if the 90 occurrences correspond to 30 different terms, UCD is 3%. And the ratio between the two gives a third reading, also published:
Repetition index = CD ÷ UCD
Here, 9 ÷ 3 = 3: each term of the subject returns three times on average. A repetition index of 2 or 3 describes a page that varies its vocabulary; an index of 5 or 6 describes a page that hammers a few words. Both can show the same CD; the repetition index tells which one was built how.
UCD has its own limit: its denominator is length, so it favours short pages. A 400-word page that places 17 different terms will show a higher UCD than a 4,000-word page that places 118, although the second covers seven times more of the subject. That is why UCD is published in the table but does not enter the score.
What the index scores: coverage and density, together
Semantic coverage answers a simple question: of all the terms that define the subject in this market, how many does the page use? Those terms form the market’s master list, extracted from the best-ranking pages for the keyword, sorted into variations, entities and LSI terms, and published with every edition. The name of every site in the ranking is on it as an entity, ours included: a page that names its competitors covers its subject better than a page that keeps quiet about them.
Coverage measures semantic depth. Its limit is known: it never penalizes length, and a very long page has a structural advantage, capped by the list. Contextual density measures efficiency: it is neutral to length, but it rewards repetition. Scored together they pull in opposite directions, and a page only scores if it is both broad and dense. That is the rule ProStar SEO applies to its own pages, which must beat the best competitor on both measures, and it has been the rule of the index since version 3.1: the content block is worth 40% of the score, split into coverage (25%) and contextual density (15%).
How the measurement is made
Every page is downloaded in its live HTML on the edition date and measured with the same script, against the same master list. The script reads the title tag, the H1 to H3 headings, link texts, ALT attributes and bold text, and takes the body text from the page’s main tag. A page rendered by JavaScript is measured on its rendered output; a page that blocks robots is measured on its displayed text, and the table says so. It is the script that validates the pages we write for our clients. Its stop-word list is published with it, and it is still English-only. Version 3.2 shipped without the French extension, so densities on French pages are computed on a denominator close to the total word count and are not comparable to the English ones. That is the oldest open item on this page.
What we aim for when we write a page
The same measures serve to write our pages and to grade everyone else’s. For a complete service page we aim for more than fifty distinct keyword variations, twenty-five to forty distinct entities per 2,000 words, more than thirty LSI terms, and at least 70% of the terms the best competitors share, while beating the best of them on both contextual density and unique density. Studying competitors does not mean copying them: their shared vocabulary is the floor, not the ceiling.
The four other signals
Coverage and density say what a page is worth. The four other signals say what the market grants the site that wrote it. They come from Semrush reports and public Google profiles, never from a third party’s composite scores.
Page-1 keywords (20%)
The number of keywords for which the domain holds a position 1 to 10 on Google, summed across the Semrush databases retained for the market. A site that ranks hundreds of pages on page 1 demonstrates a capacity the index’s single keyword does not reveal. The total goes through a logarithm before being scaled to 100, so that very large domains do not saturate the signal.
Non-branded commercial-intent keywords (15%)
The number of the domain’s positions on queries Semrush classifies as commercial or transactional, that is, searches by people comparing or buying. Queries containing the site’s name are set aside: notoriety does not inflate this signal. Same logarithmic normalization.
AI visibility (15%)
The number of queries for which the domain is cited inside a Google AI Overview. A site that sells SEO in 2026 and does not appear in generative answers is selling something it does not obtain for itself. The signal is limited to Google’s AI Overviews, the only source available through an API; ChatGPT and Perplexity are not measured, and that limit is written down.
Google reputation (10%)
The site’s Google Business Profile: the rating out of five multiplied by the logarithm of the review count, scaled to 100. Any reader can check it in two clicks. A site with no public profile gets zero on this signal, and the table says so.
The index formula
SCDI = 0.25 × Coverage + 0.15 × Contextual density + 0.20 × Page-1 keywords + 0.15 × Commercial + 0.15 × AI visibility + 0.10 × Reputation
Each signal is scaled to 100 within the group of measured sites. The three Semrush signals go through a logarithm against the group maximum, publisher included, and reputation uses its own formula. Coverage and contextual density are handled differently: since version 3.2 they are scaled against the best competitor rather than the group, and the publisher’s own pillar is capped at 100. The score is therefore relative to the market measured, and a new maximum on one signal redistributes the others’ scores; each edition says so when it happens. The weights, the databases, the master list and the number of published sites are fixed for a market at its first edition. Changing them constitutes a new version of the index, announced on this page.
Why coverage is scaled against the best competitor
Until version 3.1 every pillar was divided by the group maximum, the publisher’s page included. That created a defect we only saw once we rewrote our own page. Our page held the coverage maximum, so it was the denominator: covering more terms raised the denominator, divided every competitor’s pillar down, and left ours at the 100 it already showed. Rewriting the page could lower fourteen other scores and never raise our own. An index whose publisher is also its yardstick is measuring the wrong thing.
Version 3.2 takes the denominator out of the publisher’s hands. Coverage and density are divided by the highest figure among the competitors, and the publisher’s pillar is capped at 100 so that beating the best competitor cannot be worth more than a full pillar. In the September 7 plastic surgery edition the coverage denominator is Pasha Digital Solutions at 84 terms; our page covers 144, which is 171 before the cap and reads 100 on the table. The density denominator is Sagapixel’s 16.27%.
The cap is what keeps the change honest. Without it, a publisher that spends every term on its own list walks away with a coverage pillar of 171 out of a possible 100, and first place by construction. We computed that version: it puts us first at 93.8, which is why it is not the one published. Capped, the same page scores 75.9 and finishes fourth of fifteen, behind OuterBox, Softtrix and Helium SEO. Competitors move too, and not all downward. Pasha Digital Solutions, which sets the coverage denominator, rose from twelfth under group-maximum normalisation to eighth, because the pillar it maxes out is no longer being divided by a page it does not compete against.
One consequence is worth stating plainly, because it works against us. Our coverage pillar is already at its cap, so covering more terms adds nothing to our score. Only density still has room, and taking it from 14.4% to Sagapixel’s 16.27% would move us from 75.9 to roughly 77.6, still short of Helium SEO’s 79.7. Third place in this market is not reachable by writing. It is held by page-1 volume, commercial keyword count, AI citations and a review count of six, and only those move it.
Who is ranked, and how
The group measured for a market gathers the sites that rank a real service page in Google’s top 30, as read by Semrush, for the core keyword, plus ProStar SEO, measured with the same instrument. Directories, news articles, forum threads, blog posts and editorial rankings without a service page are excluded, with the reason: those are pages that talk about the market, not players in it. A site that ranks both an editorial ranking and a service page is kept, both positions are recorded, the table keeps the better one.
Every site in the group is measured; the table publishes the 14 best-scored and ProStar SEO, always at its real rank. When ProStar SEO is among the top 15, the table shows 15 contiguous rows; otherwise it shows the top 14, an ellipsis row, then our row. The publisher of the index never disappears from its own ranking.
Every published site is linked to the exact page measured, with a nofollow link: the reader can verify every figure, and the index passes authority to nobody. A site enters the group at the monthly edition that sees it in the top 30; it leaves after two consecutive monthly readings outside the top 30, so that a passing movement does not push sites in and out every month.
Disclosure of related properties
ProStar SEO operates or controls several sites in several markets. Before each edition, the measured group is compared with a registry of those properties. Any overlap triggers a disclosure line under the table, a mention in the page’s structured data and a note in this section. The related property is measured and ranked exactly like the others; its relationship with ProStar SEO influences no signal. Likewise, our own page is measured with the ranking it contains, which adds the names of the group’s sites as entities; the same rule applies to any page in the group that embeds a ranking, and the table publishes the coverage obtained that way without correcting it.
Updates and versions
Positions and the three Semrush signals are re-checked every week; the table, its caption and its edition badge follow. The sites’ profiles remain those of the monthly edition and say so. Every month the index is recomputed in full: group re-evaluated, pages re-downloaded and re-measured, Google profiles re-read, site profiles rewritten. Every edition is dated and archived.
The master list is fixed for the duration of a version, because it is the denominator of coverage and the numerator of the densities: replacing it silently would move every site’s figures from one month to the next. Only the names of sites entering the group are added over the editions. Every month, the index nonetheless measures its drift: the share of published sites that are new (a change of order does not count) and the gap between the fixed list and a candidate list extracted from the current pages. When a third of the published sites are new or the vocabulary gap exceeds 15%, the list is re-extracted, the edition recomputed and the change announced here as a new version.
The index has carried version number 3.1 since September 1, 2026, and version 3.2 since September 7, 2026. Version 3.0, computed on September 1 and never published, scored coverage alone at 40%; 3.1 split that block into coverage (25%) and contextual density (15%) and published the UCD column. Version 3.2 keeps those weights and changes two other things: the plastic surgery master list is rebuilt from 142 terms to 188, with the variation family re-pulled from Semrush and tiered by search volume, and coverage and density are scaled against the best competitor instead of the group maximum, with the publisher’s pillar capped. Both changes move every figure, so a 3.2 score cannot be compared with a 3.1 score, and the September 6 edition cannot be compared with the September 7 one. The French stop-word list is still not in the script.
The limits, stated plainly
Contextual density is a powerful factor, not a guarantee. Several components of the algorithm escape any outside measurement: visitor behaviour signals (click-through rate, dwell time, depth of visit, which the 2024 leak ties to the NavBoost system), brand awareness, whose effect on all of a site’s keywords has strengthened markedly since the March 2024 core update, and a site-wide quality score whose calculation remains opaque. Nor do we know exactly where relevance ends in the algorithm’s eyes: a term that seems tied to the subject to us may be treated as noise. Finally, backlink profiles are never complete, and some sites serve different content to robots. The index measures what can be measured from the outside, with the same limits for everyone.
The index does not measure the results a site obtains for its clients, only the public signals that make them plausible. Scores are comparable within a market, not across markets. Coverage tracks page length, density rewards repetition; they are scored together for that reason, and neither is perfect alone. The AI signal is limited to Google’s AI Overviews. French stop words are not yet removed from the density denominator. And the publisher of the index appears in it, sometimes at the top, which requires that the formula, the data and the exclusions all be published; they are, and any ranked site can request its raw measurements.
An example from A to Z: plastic surgery SEO
Take the keyword “plastic surgery seo services”. Semrush reads the top 30; we set aside the directories, the “Top 10 agencies” articles and the blog posts, and 14 sites remain that rank a true service page. Those 14 pages, plus the ProStar SEO page measured with them, provide the master list: 188 terms in three families, 67 keyword variations, 42 entities and 79 LSI terms, with each site’s name among the entities. The variation family is tiered by search volume against the head term’s 2,400 monthly searches, so a variation worth ten searches a month does not weigh the same as one worth three hundred.
Every page is downloaded and measured. The broadest page is ProStar SEO’s own, 4,974 useful words covering 144 of the 188 terms; the broadest competitor page is Pasha Digital Solutions at 84, and that 84 is the number every coverage figure in the table is divided by. Density ranks differently: Sagapixel sets the maximum at 16.27% on 953 words, SEO Guru Atlanta follows at 16.24%, and ProStar SEO’s page reads 14.4%, fourth of the fifteen. The shortest page, Straight North’s 451 words, covers just 25 terms; its UCD of 8.45% is the highest in the group because a short page has little room to repeat, and its repetition index of 1.56 is the lowest. Straight North also holds the fourth-best position on the keyword itself from that short page, which is the clearest reminder here that length is not the ranking variable.
Semrush completes the picture: OuterBox has 4,270 page-1 keywords and appears in 2,339 AI Overviews, both cohort maxima; Softtrix holds the commercial maximum with 5,854 non-branded commercial keywords. OuterBox’s Google Business Profile shows 4.9 on 105 reviews; Softtrix shows 4.8 on 336 reviews, the largest count in the group; SEOReseller has no profile and gets zero on that signal, which is most of the distance between ninth place and fifth.
The score assembles everything with the formula: OuterBox finishes first at 86.9, carried by its page-1 keywords and its AI visibility; ProStar SEO, measured the same way, finishes fourth of fifteen at 75.9, carried by its content, and its own entry states first what it lacks: page-1 volume and reviews. That row is measured on the published version of this page, module included, and re-measured after the module went live so it describes the page rather than the page without it. Every name in the table links to the page measured, nofollow, and every column sorts with a click. The following month the group is re-evaluated, the pages re-measured, and if a third of the published sites have changed, the master list is re-extracted and a new version announced here.
Ranking of the Best Plastic surgery SEO services by the SEO Contextual Density Index
Who actually delivers plastic surgery seo services, measured rather than self proclaimed? This table ranks the 14 true competitors holding a real service page in Google’s US top 30 for the keyword, plus ProStar SEO, publisher of the index, measured with the same instrument on the version of this page you are reading. Listicles, directories and blog posts stay out; only a genuine plastic surgery seo company earns a row. ProStar SEO ranks 4th of 15 this edition. This page carries the broadest vocabulary in the market, 144 of the 188 terms against 84 for the best competitor, and is scored as though it carried 84: from this edition the coverage and density pillars are normalised against the best competitor, and the publisher’s pillar is capped at 100. ProStar earns nothing for the 60 terms above that line. The change cost three places, and it means no competitor’s score depends on what the publisher does to its own page.
The short version
Top of the index
- OuterBox86.9
- Softtrix80.9
- Helium SEO79.7
- ProStar SEO75.9
- SEO Guru Atlanta75.0
Widest vocabulary: ProStar SEO at 144 of 188 terms, against Pasha Digital Solutions at 84 for the best competitor. Highest density: Sagapixel, 16.27%. ProStar SEO writes its pages against this same measured term list, which the reading section states plainly. It sits 4th of 15 anyway, and the four right-hand columns are why: they measure the domain, not the page.
Full table below, twelve columns, sortable.
⇅ Sortable table: click a column header to sort, again to reverse; the # column restores the ranking. Hover a header to read the metric’s definition.
| # | Agency | Pos. | SCDI | Coverage /188 | Density (CD) | UCD | Rep. | Page-1 keywords | Commercial KWs | AI visibility | Google reviews |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | OuterBox | 19 | 86.9 | 58 | 12.93% | 2.28% | 5.67 | 4,270 | 5,446 | 2,339 | 4.9 (105) |
| 2 | Softtrix | 21 | 80.9 | 64 | 6.30% | 1.82% | 3.45 | 2,977 | 5,854 | 583 | 4.8 (336) |
| 3 | Helium SEO | 20 | 79.7 | 62 | 9.80% | 2.44% | 4.02 | 2,093 | 5,452 | 603 | 5.0 (47) |
| 4 | ProStar SEO | — | 75.9 | 144 | 14.40% | 2.88% | 4.99 | 252 | 947 | 117 | 5.0 (6) |
| 5 | SEO Guru Atlanta | 15 | 75.0 | 48 | 16.24% | 3.38% | 4.81 | 650 | 1,725 | 195 | 5.0 (64) |
| 6 | Straight North | 7 | 69.7 | 25 | 13.18% | 8.45% | 1.56 | 2,578 | 3,137 | 535 | 4.4 (32) |
| 7 | Sagapixel | 4 | 69.1 | 45 | 16.27% | 6.72% | 2.42 | 452 | 814 | 80 | 4.9 (34) |
| 8 | Pasha Digital Solutions | 17 | 65.9 | 84 | 11.68% | 3.31% | 3.52 | 88 | 258 | 58 | 5.0 (2) |
| 9 | SEOReseller | 13 | 65.0 | 53 | 13.98% | 3.63% | 3.85 | 389 | 628 | 281 | no listing |
| 10 | NKP Medical | 16 | 63.3 | 51 | 10.78% | 5.39% | 2.00 | 174 | 241 | 124 | 4.2 (131) |
| 11 | JL Plastic Surgery (Dr. Jeffrey Lee) | 9 | 62.7 | 24 | 9.23% | 5.27% | 1.75 | 828 | 2,120 | 158 | 4.9 (112) |
| 12 | Plastic Surgery Studios | 6 | 61.3 | 42 | 14.05% | 5.27% | 2.67 | 217 | 142 | 28 | 5.0 (95) |
| 13 | Studio III Marketing | 3 | 58.5 | 39 | 6.31% | 1.95% | 3.23 | 339 | 605 | 67 | 5.0 (99) |
| 14 | Plastic Surgery SEO Agency (Miami) | 22 | 36.6 | 59 | 8.80% | 3.90% | 2.25 | 6 | 16 | 1 | no listing |
| 15 | PlasticSurgeonSEO.com | 10 | 34.7 | 39 | 16.16% | 4.60% | 3.51 | 0 | 14 | 5 | no listing |
Reading the table
The index rewards two different things at once, and the table shows the tension. Coverage and density grade the service page itself, both scaled against the best competitor’s figure: Pasha Digital Solutions covers 84 of 188 terms and Sagapixel compresses 16.27% density into 45 terms on a 953-word page, so those two set the denominators. Two rows read 100 on coverage, Pasha Digital Solutions at 84 terms and the publisher at 144, because the publisher’s pillar is capped there by design. The three Semrush pillars grade the whole domain instead: OuterBox, Softtrix and Helium SEO each rank thousands of keywords on page 1 with commercial visibility and AI citations to match, which is how OuterBox holds first place from 19th on Google. The repetition index tells you how each plastic surgery seo agency built its page. Straight North reaches 13.18% density across 25 terms on 451 words, a repetition index of 1.56 with unique contextual density of 8.45%. Sagapixel reaches 16.27% across 45 terms at 2.42. OuterBox reaches 12.93% across 58 terms at 5.67, the highest repetition figure in the cohort. All three approaches rank. A dash in the Pos. column means the page sits outside the Semrush top 30; “no listing” in the reviews column means no public Google Business Profile was found.
The 15 agencies in detail
1. OuterBox · SCDI 86.9
OuterBox tops the plastic surgery seo rankings without touching the top 10 on the anchor keyword, where it sits 19th. The Ohio seo agency wins on domain strength: 4,270 page-1 keywords, 5,446 commercial-intent positions and 2,339 AI Overview citations, all three at or near the cohort maximum, backed by 4.9 on 105 Google reviews. Its plastic surgery seo services page covers 58 of 188 terms at 12.93% density with the highest repetition index of the table, 5.67, and unique contextual density of 2.28%. That is a frequency-led page carried by a very strong domain, and its 11.0-point lead over the publisher comes almost entirely from the three domain pillars.
2. Softtrix · SCDI 80.9
Softtrix runs the largest commercial footprint of the cohort: 5,854 commercial-intent positions, 2,977 page-1 keywords, 4.8 on 336 Google reviews, the highest review volume of any row. It takes second on those pillars despite the weakest content row of the leading group, 64 of 188 terms at 6.3% density across 4,105 words, unique contextual density 1.82%: a long page spending its length on prose the market’s vocabulary does not reach. Ranks 21st on the anchor keyword.
3. Helium SEO · SCDI 79.7
Helium SEO takes third on a balanced profile rather than a peak: 2,093 page-1 keywords, 5,452 commercial positions, 603 AI Overview citations, 5.0 on 47 Google reviews. Its page covers 62 of 188 terms at 9.8% density across 3,480 words, repetition index 4.02. Nothing is at the cohort maximum and nothing near the floor, which is what a 79.7 looks like. Ranks 20th on the anchor keyword.
4. ProStar SEO · SCDI 75.9
ProStar SEO publishes this index and appears at its real rank, fourth of 15. This page covers 144 of 188 terms at 14.4% contextual density: the broadest vocabulary in the market, fourth of 15 on density. Its coverage pillar reads 100 because it is capped at the best competitor’s 84 terms, not because the page is exhaustive. On the same 188-term list the previous version covered 103 terms at 14.46% with a repetition index of 6.35. This one covers 144 at a repetition index of 4.99: coverage rose 41 terms and repetition fell 21% at effectively the same density, which is the trade the September 6 rewrite was built to make. The domain pillars hold the gap to the leaders: 252 page-1 keywords against OuterBox’s 4,270, 947 commercial positions, 117 AI Overview citations, 5.0 on 6 Google reviews, and no top-30 position yet for the anchor keyword.
5. SEO Guru Atlanta · SCDI 75.0
SEO Guru Atlanta posts 16.24% contextual density, second highest of the cohort and within 0.03 points of Sagapixel, on a page of 1,895 words covering 48 of 188 terms. The repetition index of 4.81 is the third highest in the table, so that density comes from saying a moderate number of things often. The domain supports it with 650 page-1 keywords, 1,725 commercial positions and 5.0 on 64 Google reviews. Fifth at 75.0, 0.91 points behind the publisher, ranking 15th on Google. Measured on displayed text in a browser.
6. Straight North · SCDI 69.7
Straight North ranks 7th on the anchor keyword, fourth-best position in the table, from a 451-word page: 25 of 188 terms at 13.18% density, repetition index 1.56, unique contextual density 8.45%, the highest here. Almost every term appears once, and the page is short enough that a small vocabulary fills it. The domain does the rest: 2,578 page-1 keywords, 3,137 commercial positions, 535 AI citations, 4.4 on 32 Google reviews. Sixth at 69.7, and the clearest case here that page length is not the ranking variable.
7. Sagapixel · SCDI 69.1
Sagapixel sets the density denominator for this edition at 16.27%, across 45 of 188 terms on 953 words, repetition index 2.42, unique contextual density 6.72%. Most of that density comes from distinct terms rather than repeats. The anchor page ranks 4th, second-best of any row here. Domain pillars are mid-cohort: 452 page-1 keywords, 814 commercial positions, 4.9 on 34 Google reviews. Measured on displayed text in a browser.
8. Pasha Digital Solutions · SCDI 65.9
Pasha Digital Solutions sets the coverage denominator for this edition: 84 of 188 terms, the broadest competitor page in the market, at 11.68% density across 3,159 words, repetition index 3.52. Every other coverage figure in the table is scaled against that 84, the publisher’s included. Everything outside the page holds it at eighth: 88 page-1 keywords, the cohort floor among domains with any presence, 258 commercial positions, 58 AI citations, 5.0 on 2 Google reviews. It ranks 17th on the anchor keyword. Under the previous cohort-max normalisation this row placed twelfth.
9. SEOReseller · SCDI 65.0
SEOReseller covers 53 of 188 terms at 13.98% density across 2,302 words, repetition index 3.85, a well-balanced content row. It scores no Google reputation at all: no public Business Profile was found for the domain, which costs it the full 10% weight of that pillar and is most of the gap between ninth place and fifth. 389 page-1 keywords, 628 commercial positions, 281 AI citations. Ranks 13th on the anchor keyword.
10. NKP Medical · SCDI 63.3
NKP Medical is a healthcare-only agency and the row reads that way: 174 page-1 keywords and 241 commercial positions, both near the cohort floor, against 4.2 on 131 Google reviews. The page covers 51 of 188 terms at 10.78% density across 1,401 words, repetition index 2.0, unique contextual density 5.39%: a page that says most things once. Tenth at 63.3, ranking 16th.
11. JL Plastic Surgery (Dr. Jeffrey Lee) · SCDI 62.7
JL Plastic Surgery, Dr. Jeffrey Lee’s practice, is the only surgeon-owned page in the cohort and ranks 9th on the anchor keyword with the second-smallest page here: 643 words, 24 of 188 terms, 9.23% density, repetition index 1.75. The domain carries it: 828 page-1 keywords, 2,120 commercial positions, 4.9 on 112 Google reviews. Eleventh at 62.7. Measured on displayed text in a browser.
12. Plastic Surgery Studios · SCDI 61.3
Plastic Surgery Studios ranks 6th on the anchor keyword, ahead of nine agencies that outscore it here, from 1,298 words covering 42 of 188 terms at 14.05% density, repetition index 2.67, unique contextual density 5.27%. The domain pillars hold the score down: 217 page-1 keywords, 142 commercial positions, 28 AI citations, against 5.0 on 95 Google reviews. A case where the page outperforms the domain behind it.
13. Studio III Marketing · SCDI 58.5
Studio III Marketing ranks 3rd on the anchor keyword, the best position of any agency here, and finishes thirteenth on the index. Its page covers 39 of 188 terms at 6.31% density across 2,716 words, the lowest density of the cohort alongside Softtrix, unique contextual density 1.95%. Domain pillars are modest: 339 page-1 keywords, 605 commercial positions, 5.0 on 99 Google reviews. The gap between 3rd on Google and 13th here is the largest inversion in the table.
14. Plastic Surgery SEO Agency (Miami) · SCDI 36.6
Plastic Surgery SEO Agency, a Miami shop on an exact-match domain, covers a genuinely broad vocabulary, 59 of 188 terms at 8.8% density across 2,091 words, third-broadest coverage in the cohort behind the publisher and Pasha. The domain has almost nothing behind it: 6 page-1 keywords, 16 commercial positions, 1 AI citation, no public Google Business Profile. Fourteenth at 36.6, ranking 22nd on the anchor keyword. Measured on displayed text in a browser.
15. PlasticSurgeonSEO.com · SCDI 34.7
PlasticSurgeonSEO.com is an exact-match domain ranking 10th on the anchor keyword with no measurable domain footprint: zero page-1 keywords in the Semrush US database, 14 commercial positions, 5 AI citations, no public Google Business Profile. The page is dense, 16.16% across 39 of 188 terms on 1,323 words, third highest in the cohort. Fifteenth at 34.7: two of six pillars read zero. It served a bot-verification wall earlier on September 6 and was measured on retry.
Frequently asked questions
What is the difference between contextual density and keyword density?
Keyword density counts how many times one phrase returns; the Cora data gives it roughly half the weight of vocabulary variety. Contextual density counts the presence of the subject’s whole vocabulary, variations, entities and LSI terms, relative to the page’s useful words.
What is the difference between CD and UCD?
CD adds up every occurrence of the subject’s terms; UCD counts each term once. Both divide by the page’s useful words. Their ratio is the repetition index.
Why does the index score coverage and CD, but not UCD?
Coverage tracks length, CD rewards repetition: scored together (25% and 15%), they correct each other. UCD favours short pages and remains a reading column.
Where do the cited correlations come from?
From Cora SEO software, which measures more than 5,000 factors across the top 100 pages of every query analysed, together with published correlation studies and ProStar SEO in-house research. The graph above is one Cora SEO 7.2 report on the query “courtier immobilier québec” (Google Canada, French, 100 results); we also hold a March 2025 reading, and the ordering of the factor categories is the same in both.
Why isn’t the number of backlinks a signal?
Its correlation with rankings has fallen to 73rd among the factors measured (−0.13), and a link count can be bought. Links still count, through their relevance, not their number.
Why can a site disappear from the table from one month to the next?
Because only the 14 best-scored sites are published, plus ProStar SEO. A site that slips from 14th to 16th is still measured but no longer displayed; it comes back as soon as it climbs. A site leaves the measured group itself only after two consecutive months outside the top 30.
Can a ranked site dispute its figures?
Yes. Every measurement comes from a Semrush report, a public Google profile or the script applied to the live page on a given date; the measurement file is provided on request.
Does the index cover other markets than the first one published?
Yes. The method applies to any keyword and any language with the same formula, the same group rules and the same disclosure registry. Each market has its own master list and its own table; scores do not compare across markets.