An SEO score is a number, usually from 0 to 100, that an audit tool uses to sum up how many technical and on-page checks a page or a site passes: title, meta description, status codes, canonicals, broken links, indexability. It isn't a Google metric: every tool calculates it with its own formula, and Google has said explicitly that it doesn't use it to rank results.
The problem is that the number gets read like a school report.
And people who read it that way usually make one of three mistakes:
- chasing 100 on checks that don't move anything, such as the alt text on a decorative icon;
- comparing scores from different tools, as if a 72 from Semrush and a 72 from Lighthouse measured the same thing;
- taking a high score as the all-clear, then not understanding why traffic stays flat.
Two projects we work on show how little the score explains about results.
Macropix, a Milan-based maker of custom LED screens, went from position 88 to 2 for "monitor pubblicitario" (advertising display) on Google Italy between 2020 and 2025, and today holds 25.55% share of voice on the LED wall cluster, ahead of Amazon at 18.08%. As we explain in our guide to the SEO audit, the lever that moved its rankings wasn't technical.
Sunscape, an Italian solar farm developer, went from 1 to 38 keywords on page one out of 143 in the six months from March 2026, with monthly impressions rising from 329 to 5,809. The work was architecture: one page per region.
And the audit score, in all this?
In neither case is it the number that tells you what happened. To measure that rather than just say it, on 9 October 2026 we calculated the SEO score of the page-one results for three searches on Google Italy. The numbers are further down.
In this article:
- what the SEO score measures and what Google has said about it;
- the formulas of five tools (Lighthouse, Ahrefs, Semrush, Seobility, Siteimprove) side by side;
- why a noindexed page scores about 69 out of 100 on Lighthouse;
- the test on 27 organic positions on Google Italy;
- what a good score is and what to fix first.
Let's start.
What an SEO score is (and why Google doesn't have one)
An SEO score is a summary index built by audit software: the tool crawls a page or a whole site, runs a list of checks and turns the ratio of passed to failed checks into a number.
Google doesn't publish a score like that, and doesn't use one.
In the 2023 episode of Ask Googlebot, John Mueller, Search Advocate at Google, put it like this: "no, Google does not use scores from third-party SEO tools for search".
In the same video he adds a sentence that gets quoted less often: "Transparently calculated scores can be useful to estimate your website's standing or to point out actual issues".
Everything else rests on that distinction. The score is useful if you know which checks it contains. If you don't, it's just a number.
One clarification to avoid confusion: Moz's Domain Authority, Ahrefs' Domain Rating and Semrush's Authority Score are a different family of scores. They measure links, not the technical state of the page, and we compared them in our guide to Domain Authority. Here we're only talking about health and on-page optimisation scores.
How five tools calculate the SEO score (only two publish the full formula)
Every tool calls something different a "score". This is what each one's official documentation says, checked on 9 October 2026.
| Tool | Name of the score | How it's calculated | Backlinks included |
|---|---|---|---|
| Lighthouse (Google Chrome) | SEO | 10 pass/fail checks, weighted; "is-crawlable" weighs about 4 times the others (source code) | No |
| Ahrefs Site Audit | Health Score | Internal URLs without errors divided by total internal URLs, times 100 (Ahrefs Academy) | No |
| Semrush Site Audit | Site Health | Errors and warnings relative to the number of checks run; errors weigh more than warnings; exact formula not published (Semrush KB) | No |
| Seobility | SEO score | Over 200 factors, weights not published; 80% and above is a "solid base", below 30% "critical problems" (Seobility) | Mentions "external signals" among the factors, without detail |
| Siteimprove | SEO Score | 4 subcategories: Technical (high priority), Content and Mobile (medium), UX (low); numerical weights not published (Siteimprove) | No, explicitly excluded |
The practical consequences show straight away.
On Ahrefs a single template error can push the Health Score close to zero. If a broken link in the footer appears on every page, no internal URL is "error-free" and the ratio collapses, even if the rest of the site is in order. On Semrush, which relates errors and warnings to the checks run, the same error has a different effect.
Lighthouse looks at one page at a time, Ahrefs and Semrush at the whole crawled site. A 90 on Lighthouse and a Health Score of 90 describe two different things.
Three tools out of five don't publish their weights. That means when the score goes up by 4 points you don't know which fix earned them, unless you reread the report check by check. What these tools cost, and which ones you actually need, is covered in our guide to the best SEO tools.
Lighthouse: why a page Google can't see scores 69 out of 100 (the maths from the code)
Lighthouse is the only one of the five with a fully public formula, because it's open source. Its SEO category contains 10 weighted checks and one manual check, structured data, with zero weight.
Nine checks weigh 1: title, meta description, HTTP status code, link text, crawlable links, robots.txt, image alt, hreflang and canonical.
The tenth, is-crawlable, checks that the page isn't blocked from indexing by a noindex or by robots.txt. It weighs 93/23, or about 4.04. The comment in Lighthouse's configuration file explains why: it has to be worth at least 31% of the score, so that if it fails, the category fails.
Let's do the sum.
| Lighthouse SEO check | Weight | Points out of 100 |
|---|---|---|
| is-crawlable (page can be indexed) | 4.04 | 31.0 |
| Each of the other nine (title, meta description, status code, link text, crawlable links, robots.txt, alt, hreflang, canonical) | 1 | 7.7 |
| Structured data (manual) | 0 | 0 |
| Total | 13.04 | 100 |
A page that passes everything except is-crawlable, in other words a page Google can't index, scores about 69 out of 100.
On the colour scale the Chrome documentation describes for scores (red from 0 to 49, orange from 50 to 89, green from 90 to 100), 69 is orange: "needs improvement".
So a page that's out of Google gets the same colour as one with a few missing alts?
Yes. An indexable page with no title and images without alt scores about 85: orange too, with a problem of a completely different kind.
One telling detail: the scoring guide in the Lighthouse repository still says the SEO checks are equally weighted and each is worth "about 8 points". The configuration file, which is what the tool actually runs, says otherwise. Even a score's documentation needs checking.
What about speed? Lighthouse's SEO category doesn't count it: its description warns that there are many other factors it doesn't assess, Core Web Vitals included.
Your audit says "needs improvement" and you don't know if it's serious?
A page that's out of the index and one with two missing alts can show the same colour. We'll go through which errors on your site affect pages that bring in clicks, and which can be left where they are.
Book a strategy consultationThe test: SEO score and rankings on three Italian searches (correlation -0.10)
Across 18 page-one results on Google Italy, the SEO score showed no relationship with ranking position: a Spearman correlation of -0.10, with a p-value of 0.68.
The method. On 9 October 2026 we pulled the Google Italy mobile SERP with DataForSEO for three different kinds of search, in Italian: a local one ("commercialista milano", accountant Milan), an informational one ("come fare il curriculum", how to write a CV) and an e-commerce one ("materasso memory", memory foam mattress). For each organic result we calculated DataForSEO's onpage score, a 0 to 100 score based on on-page checks, and for the third search the Lighthouse SEO score as well. We ran it on the Italian market, where most of our clients are; it would take around thirty separate page analyses to repeat it on Google UK, so we've kept the Italian results as they are.
| Search | Organic positions | Scores read | Average, positions 1-3 | Average, positions 4-10 | Minimum and maximum | Position/score correlation |
|---|---|---|---|---|---|---|
| commercialista milano | 10 | 6 | 96.53 | 95.52 | 94.15 - 98.90 | 0.15 |
| come fare il curriculum | 7 | 5 | 97.07 | 94.63 | 94.15 - 97.07 | -0.87 |
| materasso memory | 10 | 7 | 91.04 | 92.59 | 87.93 - 97.44 | 0.05 |
| Total | 27 | 18 | 94.88 | 94.08 | 87.93 - 98.90 | -0.10 |
Five things the test tells us.
One: the top three positions score less than a point higher on average. 94.88 against 94.08. On the CV search the correlation is stronger (-0.87, meaning better positions with higher scores), but across 5 pages and with a p-value of 0.054. We wouldn't build anything on it.
Two: the top result for "materasso memory" has the lowest score in the test. materassiematerassi.it, first and fifth with the same URL, scores 87.93: a 6-character title, no H1, no meta description. Lighthouse gives it 92.
Three: all the scores sit within 11 points. From a minimum of 87.93 to a maximum of 98.90. A 0 to 100 scale used within an 11-point band tells you little about pages that already rank, because almost all of them have a title, a canonical and HTTPS. We'd found something similar, on a single search, in our guide to technical SEO: the page with the lowest score was fourth.
Four: the tool couldn't even read more than a third of the positions. 10 positions out of 27, or 9 different URLs, returned a 403 error to the crawler: PagineGialle, PagineBianche, ProntoPro, Trovaprezzi, inmaterassi.it and the top result for the CV search, ilcvperfetto.it. For the tool these pages have no score. For Google they do: they're on page one.
Five: two tools, two verdicts on the same page.
| Position for "materasso memory" | Site | DataForSEO onpage score | Lighthouse SEO |
|---|---|---|---|
| 1 | materassiematerassi.it | 87.93 | 92 |
| 2 | inmaterassi.it | not read (403) | 40 |
| 3 | goldflexmaterassi.it | 94.15 | 92 |
| 4 | olimpo-flex.com | 97.44 | 100 |
| 7 | my-personaltrainer.it | 97.07 | 85 |
| 8 | emma-materasso.it | 92.32 | 100 |
| 9 | trovaprezzi.it | not read (403) | not read |
| 10 | lettiematerassi.com | 88.19 | 100 |
lettiematerassi.com is second to last for DataForSEO and scores 100 on Lighthouse. my-personaltrainer.it is second for DataForSEO and the lowest on Lighthouse among the pages read in full.
And inmaterassi.it, second on Google, scores 40 on Lighthouse. We don't know whether Lighthouse read the real page or an anti-bot protection page, given that the other tool got a 403 at the same address. The score measures what the server shows the tool, which isn't necessarily what it shows Googlebot.
And what does your own page score?
In the same test, our Italian guide on how long SEO takes to show results scores 100. That means the title, meta and canonical are in order. It doesn't say anything else.
What the SEO score sees well (four cases where it's worth a look)
The score is reliable when the error is binary and repeats across many pages. That's the use Mueller describes: transparent scores to find real problems.
- A noindex left in place after a release. On Lighthouse it's worth 31 points on its own, and on Google it's worth the whole page. If a page disappears from search, the first check is in our guide to indexing.
- A migration or a redesign. Compare the score from the day before and the day after go-live: a sharp drop within 24 hours almost always means redirects, canonicals or robots.txt have changed. What usually breaks is in our guide to website redesign and SEO.
- Template errors. Duplicate titles or wrong canonicals across every product page: one error, hundreds of URLs. How to read these tags is explained in our guide to meta tags.
- The trend over time. Same tool, same configuration, month after month. The absolute value says little, the change says a lot. The schedules we use are in our SEO checklist.
What the SEO score doesn't see (links, intent, competitors)
Almost nothing that decides your position on a competitive search goes into the score.
- Links. Siteimprove says so plainly: its SEO Score doesn't take the backlink profile into account. Ahrefs and Semrush keep it in separate metrics.
- Search intent. The tool checks that the title exists and is the right length, not that it answers what people are looking for. If it doesn't, Google rewrites it: we covered this in our guide on when Google ignores the meta title.
- Competitors. On Google UK on 10 October 2026, for "accountant london", 5 of the 10 organic results were directories and portals (Unbiased, ICAEW's Find a Chartered Accountant, Yell, FreeAgent, The Manifest), with a further carousel of Yell, Bark, ThreeBestRated, Cylex and Yelp above the map. No audit of your site changes that: it's one of the SEO challenges no audit solves.
- Bots. In the test 9 URLs blocked the tool's crawler. Who gets through and who doesn't is decided in robots.txt and the firewall, as we explain in our guide to AI crawlers.
- Google's own data. Impressions, clicks and indexed pages are in Google Search Console, not in an external audit.
What is a good SEO score? (It depends on the tool, and 100 isn't the goal)
A good SEO score is one without blocking errors: no accidental noindex, no 4xx or 5xx pages linked from the site, no canonical pointing to the wrong URL. The number that matches this condition varies from tool to tool.
| Tool | "Good" threshold given by the vendor |
|---|---|
| Lighthouse | 90-100 (green) |
| Seobility | 80% and above, "solid SEO base" |
| Ahrefs | above 80 "good", above 90 "excellent" (Ahrefs blog) |
Bear in mind the Lighthouse case above: a noindexed page sits at 69, below the green threshold but far from red. The threshold alone doesn't tell a serious problem from a cosmetic one.
The rule we use: zero blocking errors, then stop looking at the number and work on content, architecture and links.
How to improve your SEO score without working for the number (the order we follow)
You improve it by fixing the errors that take the page out of Google first, and the ones that only cost points last.
- Indexability: noindex, robots.txt, status codes. On Lighthouse they're worth 31 + 7.7 + 7.7 points; on Google they're the condition for existing at all.
- Canonicals and redirects: a wrong canonical sends the signals to the wrong page.
- Broken internal links and link text: they're also how Google discovers pages.
- Title and meta description: they matter for the click, but the top result for "materasso memory" had a 6-character title.
- Speed: Lighthouse's SEO category doesn't count it, Google does, through Core Web Vitals.
- Alt text on decorative images: last.
Then there's the filter no tool applies on its own: cross-check the list of errors against the clicks in Search Console. An error on a page with zero impressions in 28 days can wait. The same error on the page that brings in the most enquiries needs fixing today.
The full procedure, check by check, is in our guide to the SEO audit. If you'd rather hand it to someone, it's the first step of our SEO service.
FAQs
No. John Mueller, Search Advocate at Google, repeated it in 2023 in an episode of Ask Googlebot: Google does not use scores from third-party SEO tools for search. Lighthouse, even though it is a Google project, is no exception.
It depends on the tool and on what sits behind the number. On Lighthouse green starts at 90, Seobility gives 80% as a solid base, and Ahrefs considers a Health Score above 80 good. The criterion we use is zero blocking errors, because on Lighthouse even a noindexed page scores about 69.
Yes, and it happens often. In our test on 9 October 2026, lettiematerassi.com scored 88.19 on DataForSEO and 100 on Lighthouse: the two tools run different checks with different weights.
No. Across the 18 page-one results we measured on Google Italy, the correlation between score and position was -0.10, and the top result for "materasso memory" (memory foam mattress) had the lowest score in the test, 87.93.
It depends on the tool. Lighthouse, the Ahrefs Health Score, Semrush Site Health and Siteimprove do not; Seobility mentions external signals in general terms. Links are measured with separate metrics such as Domain Rating or Authority Score.
It depends on how often you release changes. Always the day after a migration, a redesign or a theme update; otherwise one crawl a month is enough to see the trend.
The SEO score in five lines (and the number to look at instead)
The SEO score is a tool metric, not a Google one.
Every tool calculates it its own way: Ahrefs counts URLs without errors, Lighthouse weighs 10 checks.
On Lighthouse a noindexed page scores about 69.
Across 18 page-one results on Google Italy, the correlation with ranking position was -0.10.
And the tool couldn't even read more than a third of the positions.
The score doesn't appear in the reports we send clients. In its place we put a number no tool calculates on its own: how many pages with a blocking error have had at least one click in the last 28 days. If it's zero, the score can wait. If it isn't, no 95 in an audit will protect you.
Until next time!
