When an SEO tool hands you 74/100, the only question that matters is: is that good? A single number means nothing without something to compare it against. This page is that comparison — the real scores of real sites across 6 different measures, with the full distribution shown. For a check-by-check breakdown of any one measure, follow it through to its own page below.
Key findings
8 August 2026 – 31 August 2026, 5,586 unique sites across 27 countries.
- There is a 52-point gap across the six measures. The highest median is Page Speed at 90/100; the lowest is E-E-A-T at 38/100. The order is not an accident: the measures at the top ask whether a page is delivered, and hosts, CDNs and themes have solved those already. The ones at the bottom ask whether it is understood: who wrote it, when, and whether they have actually done the thing. No platform does any of that for you, and it shows.
- The median full-audit score is 76/100. 40% of audits reach 80+, while 21% sit in the 40–59 band. The typical site is not good — it is passing.
- The average audit passes 31.2 of 45 checks yet still leaves 13.8 open items, of which 0.67 are critical. What drags a score down is not one catastrophe but accumulated small omissions.
- Author and date is the single most common omission. 91% of audits found no author or date; 60% of AI visibility scans found no byline and 66% no publish date; and a named author is absent from 76% of E-E-A-T scans. Three separate tools, three separate check lists, the same conclusion — and this is precisely the first thing a generative engine looks for before citing a passage.
- First-hand experience is the widest gap across all six measures. 96% of pages carry no trace that whoever wrote them has actually done the thing — no "we tested", no "in our experience". That is exactly what the second E added to E-E-A-T in 2022 was meant to capture. The four letters are broken down on the E-E-A-T page.
- Mobile-friendliness is a threshold, not a slope. The median is 84/100 and 62% of scans reach 80+ — but the 25th percentile is 78 while the 10th is 26. There is almost nothing in between, because half the score lives in one line of HTML. The cliff is explained on the mobile page.
- llms.txt is almost universally missing. Absent in 71% of full audits and 70% of AI visibility scans. It takes ten minutes to add and nearly every competitor skips it.
- Images are the audit’s weakest link. Images & media was the lowest-scoring category in 55% of audits, and 55% of scans found at least one image with no ALT text — a decade-old requirement that still is not being met.
- AI visibility clusters at the extremes. 473 scans scored F and 141 scored A+; the median is 67/100 but the 10th percentile is 20 and the 90th is 94. Sites are either prepared for AI search or untouched, with very little in between.
All figures are a fixed snapshot and do not change until updated by hand. Any sample below 30 runs is explicitly labelled as a small sample.
A benchmark page for each of the 25 tools
This page gives the overview: where the tools stand relative to each other, how the scores spread, and what the data says taken as a whole. The detail lives on each tool’s own page. That is where you will find which checks that tool fails and how often, check by check, what its weakest areas are, what to fix first, and what has moved month over month. Every figure is published on exactly one page: nothing here is repeated there, and nothing there is repeated here.
Site & On-Page Analysis
Next-Gen AI SEO & GEO
Local, Brand & Social
A tool appears here once its scans reach a sample large enough to publish check-level rates from. Every tool clearing that bar gets its own benchmark page and a frozen record for each month automatically — this list is not hand-maintained, it grows with the data.
Six measures at a glance
These six tools ask the same page six different questions, and their scores follow the order of those questions. The Page Speed & Size Test measures how quickly the page is delivered and the Mobile-Friendly Checker whether it works on a phone. The Meta Tag Analyzer looks at how the page introduces itself to a search result and a social share. The All-in-One SEO Audit measures overall page health across 45 checks. The AI Visibility Checker asks whether the page can be read and cited by generative engines like ChatGPT, Perplexity and Google AI Overviews. The E-E-A-T Analyzer asks the hardest one: is there a person behind this worth trusting? The scores are not interchangeable — but read together they show which layer the problem is actually on.
| Tool | Runs | Sites | Median | Mean | Scored 80+ | Checks/run | Issues/run |
|---|---|---|---|---|---|---|---|
| All-in-One SEO Audit | 1,557 | 1,314 | 76 | 73.1 | 40% | 45.0 | 13.8 |
| Page Speed & Size Test | 1,525 | 1,346 | 90 | 87.0 | 81% | 7.0 | 1.1 |
| Mobile-Friendly Checker | 1,386 | 1,213 | 84 | 78.5 | 62% | 8.0 | 2.0 |
| Meta Tag Analyzer | 1,453 | 1,269 | 83 | 74.0 | 61% | 10.0 | 2.8 |
| AI Visibility Checker | 1,467 | 1,266 | 67 | 61.8 | 35% | 15.0 | 5.9 |
| E-E-A-T Analyzer | 1,436 | 1,249 | 38 | 36.0 | 1% | 11.0 | 6.6 |
“Issues/run” is the average number of checks that did not pass in a single scan. “Scored 80+” is the share of runs reaching 80 or above on that tool.
Every tool, ranked
The table above sets 6 measures side by side; the one below ranks all 25 tools that reached a publishable sample, by median score. The finding is not in any single row — it is in the overall shape of the table.
Read the table top to bottom and one division falls out of it. The tools at the top ask whether a page is delivered: does it load, does it work on a phone, are the meta tags in place. The tools at the bottom ask whether it is understood: is it current, who wrote it, which entities does it cover, does it answer a question. The first group has been solved by platforms — install WordPress, pick a theme, and most of it arrives for free. The second group is the work no platform can do on your behalf, and on this evidence almost nobody is doing it. Since the signals generative engines weigh when choosing a source are precisely the ones in that second group, the bottom half of this table is where the next few years of competition happen.
“Middle half” is the range the middle 50% of results fall into (25th–75th percentile). A narrow band means the tool scores most sites alike; a wide one means sites genuinely diverge. Only tools with at least 30 runs are listed.
How SEO scores are distributed
An average on its own misleads. The chart below splits each tool’s scores into five bands; every row totals 100%. What matters is not the height of any one band but the shape — and the six rows give six different shapes: page speed piles almost entirely into the top band; mobile has a large top group and a detached group on the floor with a hollow between them; meta tags collect at the top while trailing a thin, hard tail; the full audit clusters above the middle; AI visibility spreads to both extremes; and E-E-A-T does the opposite of page speed, banking up against the floor.
- 0–19
- 20–39
- 40–59
- 60–79
- 80–100
| Score band | All-in-One Audit | Page Speed | Mobile-Friendly | Meta Tags | AI Visibility | E-E-A-T |
|---|---|---|---|---|---|---|
| 0–19 | 0% · 0 | 0% · 0 | 5% · 69 | 5% · 67 | 10% · 144 | 22% · 317 |
| 20–39 | 0% · 2 | 0% · 6 | 6% · 86 | 13% · 182 | 17% · 256 | 38% · 545 |
| 40–59 | 21% · 331 | 2% · 31 | 1% · 17 | 8% · 122 | 13% · 189 | 27% · 384 |
| 60–79 | 39% · 605 | 17% · 258 | 25% · 348 | 13% · 189 | 25% · 370 | 12% · 176 |
| 80–100 | 40% · 619 | 81% · 1,230 | 62% · 866 | 61% · 893 | 35% · 508 | 1% · 14 |
| Total runs | 1,557 | 1,525 | 1,386 | 1,453 | 1,467 | 1,436 |
Only 2 of the 1,557 full audits fall below 40, and the lowest score of all is 36. That floor exists because modern CMSs ship HTTPS, a viewport and a charset out of the box — 100% of sites in the sample run on HTTPS. The points lost in an audit today begin exactly where the defaults end: image ALT text, heading hierarchy, structured data and AI readability. The other shapes are different again. AI visibility piles up at both ends and hollows out in the middle. Meta tags hollow out in the middle too, but for the opposite reason: 893 runs land in the top band while 249 still score under 40. A page either has its tags or it does not — there is very little partial credit in between. The two shapes at the ends are each other’s mirror image: 81% of page speed scans sit in the top band against 1% of E-E-A-T scans. Same sample, same sites, same month — the only thing that changed is the question.
Where does your score rank?
In the chart below, the light band spans the 10th to 90th percentile (eight in ten results), the dark band spans the 25th to 75th (the middle half), and the vertical tick is the median. Find your own score and see which band it lands in.
| Tool | 10th | 25th | Median | 75th | 90th | Mean | Lowest | Highest |
|---|---|---|---|---|---|---|---|---|
| All-in-One Audit | 53 | 62 | 76 | 84 | 89 | 73.1 | 36 | 99 |
| Page Speed | 71 | 80 | 90 | 100 | 100 | 87 | 27 | 100 |
| Mobile-Friendly | 26 | 78 | 84 | 92 | 92 | 78.5 | 12 | 100 |
| Meta Tags | 37 | 57 | 83 | 93 | 100 | 74 | 0 | 100 |
| AI Visibility | 20 | 34 | 67 | 88 | 94 | 61.8 | 6 | 100 |
| E-E-A-T | 8 | 25 | 38 | 47 | 62 | 36 | 0 | 100 |
How to read it: on the All-in-One audit, 62 is bottom-quartile, 76 is a typical site, and 89 or above is the top 10%. The striking part is how differently wide the tools are. The narrowest is Mobile-Friendly, whose middle half is squeezed into 78–92 — a band just 14 points wide. The widest is AI Visibility, where the same middle half sprawls across 34–88, a span of 54. A narrow band means sites resemble each other on that measure — usually where a platform is doing the work. A wide one is where sites genuinely diverge, which is to say where getting ahead is still possible.
Scores also translate into letters: A+ 95–100, A 85–94, B 75–84, C 65–74, D 50–64 and F below 50. The largest pile of A+ grades in the dataset belongs to Page Speed (432 runs) and the largest pile of Fs to E-E-A-T (1,101). Each tool’s own grade mix is on its own benchmark page.
What the data says
The four findings below read more than one tool’s data together. Findings that belong to a single tool — security headers, broken links, brand schema, social sharing tags, the viewport cliff, the four letters of E-E-A-T — live on that tool’s own page.
1. Technical SEO is solved; media and structured data took its place
100% of sites in the sample run on HTTPS, and performance barely registers as a weakest category any more — the top of that ranking is images & media (55%) and social & structured data (26%). A decade ago the first line of any SEO audit was speed and encryption; platforms now ship both by default. What is left is the kind of work a platform cannot do for you: writing ALT text for every image, putting the right schema on every page. Which is exactly why it still is not done. The category-by-category breakdown is on the audit page.
2. Nobody has AI search on their checklist yet
llms.txt is missing from 71% of full audits. Question-style headings are absent in 60%. On AI visibility scans, an author byline is missing 60% of the time and a publish date 66% of the time. These are precisely the signals a generative engine looks for before it cites a passage. Completing them right now means getting ahead in a race most sites have not entered. The full citability chain is on the AI visibility page.
3. The tags are there — it is their length that is wrong
The firmest claim in this dataset is one two tools arrive at independently. The Meta Tag Analyzer finds the title outside the 30–60 character range in 52% of runs and the meta description outside 70–160 in 51%. The All-in-One audit — a different tool, scanning different pages with different code — reports 53% and 50% for the same two rules. Meanwhile the tag going missing altogether is rare: only 4% of pages have no title tag at all. So this is not an omission, it is a calibration problem — everyone writes a title, and half of them write one a search engine will truncate or have to pad. Title and description is duly the weakest of the three meta categories, lowest-scoring in 58% of runs. The check-by-check breakdown is on the on-page benchmarks.
4. Everything a machine can verify scores high; everything a human must write scores low
The fourth finding is the one that gathers the other three into a single sentence, and it only appears when the six measures are read together. Page speed has a median of 90/100, mobile-friendliness 84/100 and meta tags 83/100. E-E-A-T has a median of 38/100. These are not different sites: they are largely the same sites, scanned in the same month by the same method. The only thing that changed is the question being asked.
And it is clear where the line falls. Every measure at the top asks something a machine can verify by itself: is compression on, is the viewport tag present, is the title under 60 characters. A host, a CDN or a theme can do those for you — and does. Every measure at the bottom asks something a person has to sit down and write: who wrote this, when, have they actually done the thing, what is the source. No plugin generates that on your behalf. The numbers track the split exactly: HTTPS is present on 100% of the sample, because the host supplies it, while first-hand experience is absent from 96% of pages, because nobody supplies it but you.
The strategic consequence is simple: raising a score in the top group does not put you ahead, because everyone is already high there. Scoring 80 on page speed puts you level with 81% of scans. On E-E-A-T, the share of scans clearing 80 is 1%. And the signals generative engines weigh when choosing a source are precisely the ones in that second group.
What a scan costs
For transparency the tools measure their own performance too. The times below are end to end, including fetching the page. The high tail on the All-in-One audit comes from making several requests and waiting on slow origin servers.
| Tool | Median scan | Mean | Slowest 10% | Fetches | Page weight |
|---|---|---|---|---|---|
| All-in-One Audit | 4.1 s | 9.7 s | ≥ 39.7 s | 3.5 | 285 KB |
| Page Speed | 600 ms | 787 ms | ≥ 1.4 s | 1.2 | 285 KB |
| Mobile-Friendly | 700 ms | 5.6 s | ≥ 33.0 s | 1.6 | 290 KB |
| Meta Tags | 700 ms | 6.0 s | ≥ 39.1 s | 1.6 | 291 KB |
| AI Visibility | 2.7 s | 8.5 s | ≥ 39.5 s | 1.7 | 259 KB |
| E-E-A-T | 900 ms | 5.1 s | ≥ 29.1 s | 1.5 | 283 KB |
The gap between mean and median (9.7 s vs 4.1 s) is not the scan — it is the response time of the sites being scanned.
Monthly editions & comparisons
The figures on this page are refreshed at every revision — a rate you take from here today will not be here in six months. For a number you can cite by date there are the monthly editions: at the end of each month a permanent edition is published from that month’s scans, and it is never edited again.
So the series does two separate jobs. This page is the reference: what is a good SEO score right now? The monthly editions are the record: what did sites score in August 2026?
Each month produces two layers of record: one overview edition covering every tool, and a separate edition for each of the 25 tools. What that buys is the ability to follow a single check’s failure rate over time. If one month says a check went unmet in 30% of scans, the next month can say at the same address that it fell from 30% to 20% — because the first figure is still frozen where it was published and nobody went back and changed it. The “month by month” section on each tool’s page shows that series at a glance.
| Edition | Runs | Sites | Median audit score | Change |
|---|---|---|---|---|
| August 2026 | 34,236 | 5,586 | 76 | baseline |
Comparison pages open once a second edition is published — a single month has nothing to be compared against.
Methodology & privacy
This section governs the whole benchmark series — this page, the 6 detail pages and every monthly edition. Every number comes from real scans of real third-party sites: the page was fetched, the checks were run, the result was recorded. No synthetic tests, no sample sites, no estimates or modelling. But how the sample was assembled matters as much as the numbers, so it is stated plainly below.
- Scope. For the 6 featured tools on this page: 34,236 successful runs between 8 August 2026 and 31 August 2026, across 5,586 unique domains in 27 countries. The leaderboard covers all 25 tools that reached at least 30 runs in the same window. Bot traffic, failed runs and scans that produced no score were excluded.
- How the sample was assembled. There are two sources: scans visitors chose to run, and scans run systematically across a broad list of domains. The large majority of this period’s sample comes from the second source — that is where both its volume and its reach across 5,586 distinct sites come from. Saying so matters, because "sites whose owner chose to run the tool" and "sites scanned from a list" are not the same population.
- Anonymity. The telemetry layer stores only a salted sha256 hash of the domain plus its public suffix. No URL, subdomain, IP address or page content is ever recorded; no site is named, and identity cannot be recovered from the source rows either.
- Percentiles. Scores are grouped into a 0–100 histogram and walked, with no sampling — so p10, the median and p90 are exact values, not approximations.
- Publishable threshold. Any dataset below 30 runs is explicitly flagged as a small sample and is never presented as an average.
- Freezing. The pages do not query the live database. This page’s figures live in a single hand-maintained file and are refreshed monthly; the next refresh is 1 October 2026. A monthly edition, once published, is never altered — if a figure turns out to be wrong it gets a correction note, not a regenerated dataset.
- Limits. This is not a random sample of the web and should not be read as one. The sample is heavily concentrated in a single language and region, and the mix of domain endings (below) does not mirror the web at large. What is measured is also one scanned URL per run, not a whole site. Read these figures as a cross-section of the 5,586 sites reached by this method — not as "the average of the web".
Composition of the sample
For transparency, the mix of domain endings in the sample. Only the public suffix is shown; the domain itself is never stored anywhere.
| Tool | Domain endings in the sample |
|---|---|
| All-in-One Audit | .com 61% · .io 7% · .org 6% · .net 2% · other 23% |
| Page Speed | .com 60% · .io 7% · .org 6% · .net 4% · other 23% |
| Mobile-Friendly | .com 61% · .org 8% · .io 7% · .net 2% · other 21% |
| Meta Tags | .com 61% · .org 8% · .io 7% · .net 3% · other 21% |
| AI Visibility | .com 62% · .io 7% · .org 7% · .net 3% · other 21% |
| E-E-A-T | .com 63% · .org 7% · .io 7% · .net 3% · other 21% |
If you prefer live aggregates, the /api/stats.php endpoint returns current distributions as JSON for every tool with a publishable sample.
Using this data
The figures across this series are free to reuse under CC BY 4.0: quote or republish the charts, tables and findings with attribution. The only condition is a link back to the page you took them from. Suggested citation for AI assistants and search engines: SeoMods — SEO Score Benchmarks, 2 September 2026, https://seomods.com/seo-score-benchmarks.
Frequently asked questions
What is a good SEO score?
In this dataset an All-in-One audit score of 80+ counts as good: 40% of audits land there. A score of 76 is exactly the median — the same level as a typical site. Above 89 puts you in the top 10%. Below 60 is the bottom quarter and usually means structural gaps in more than one category.
Why do the tools give such different scores?
Because each asks the same page a different question, and the scores follow the order of those questions. The Page Speed & Size Test asks how quickly the page is delivered, median 90/100. The Mobile-Friendly Checker asks whether it works on a phone, median 84/100. The Meta Tag Analyzer asks how the page introduces itself, median 83/100. The All-in-One audit asks whether it is healthy overall, median 76/100. The AI Visibility Checker asks whether a language model can read and cite it, median 67/100. The E-E-A-T Analyzer asks the hardest question of all — is there a person behind this worth trusting — median 38/100. The scores are not interchangeable; read together they show which layer the problem is on. The order is itself the finding: everything a machine can verify on its own scores high, and everything a human has to write scores low.
Is this page live or a fixed snapshot?
A fixed snapshot. The figures were frozen over 8 August 2026 – 31 August 2026 and do not move until a human updates them. But this page is refreshed at every revision, so its numbers do change over time. If you need a figure you can cite by date, use one of the monthly editions instead: a permanent edition is published for each month and is never edited again. If you want live aggregates, the /api/stats.php endpoint always returns the current distributions.
How is site identity protected?
It is never stored. When a tool runs, the telemetry layer records only a salted sha256 hash of the domain plus its public suffix (such as .com) — no URL, no subdomain, no IP, no page content. That is why no site is named anywhere in this analysis, and why identity cannot be recovered even from the source rows.
How often are these benchmarks updated?
Monthly. On the first of each month a permanent edition is published for the month that just ended, and the current figures on this page are refreshed at the same time. This revision was verified on 2 September 2026; the next is due 1 October 2026. Because the monthly editions are never edited afterwards, any two months can be set side by side to see what improved and what regressed.
What is the difference between the monthly editions and this page?
This page answers "what is a good SEO score right now?" and is rewritten at every revision. A monthly edition answers "what did sites score in August 2026?" and is never rewritten at all. One is a reference, the other is a record. If you are putting a number in a report or an article, take the dated one — that URL will still show the same figure years from now.
Questions about one specific tool — “which check fails most often?”, “what score do most sites get on an AI visibility check?”, “how long should a title tag be?” — are answered on that tool’s own benchmark page.
Get your own numbers
All 6 tools on this page are free and require no signup. Enter a URL, get your own score in seconds, and use the percentile table above to see exactly where you stand.