The State of Web Readability 2026

Q: What reading level is most web content written at?

In this crawl, the median page across all slices sits where the summary tiles say, with the top-1,000 homepages and SaaS pages at one end and government and university pages at the other. Prior studies of pages that rank in Google put the typical page around grade 11 to 13 on the classic formulas; this study reads whole sites' public pages rather than search results, scores them with the same extractor and the same formulas, and adds the Orwellix four-dimensional grade with the hard sentences counted.

Q: What grade level should my website be?

The Nielsen Norman Group's advice, which most style guides repeat, is 8th grade for a broad consumer audience, 6th grade on landing pages for lower-literacy readers, and 10th to 12th for expert readers. Google states no target and its rater guidelines do not mention reading level. Use the slice closest to your audience in the table as a benchmark, and the hard-sentence share rather than the grade alone.

Q: Are websites getting harder or easier to read?

This is the first edition, so the trend will come from next year's crawl of the same slices with the same method. The only long-run measurements we could open are for privacy policies, which got longer and about a grade harder between 2000 and 2019, and for government sites in 2003, which read at the 11th grade on Flesch-Kincaid.

Q: Why score homepages? They are mostly headlines.

Because that is what people land on. The extractor keeps only sentence-shaped paragraphs of at least eight words, drops menus, cookie banners and footers, and discards any page with fewer than 80 words of prose, so the score describes the sentences a visitor actually reads. The word count per page is in the table, and the per-page CSV lets you filter to longer pages.

Q: How does the Orwellix grade differ from Flesch-Kincaid here?

Flesch-Kincaid counts words per sentence and syllables per word. Homepage copy is short-sentenced and is read as easy by the formula even when it is dense with product jargon. Orwellix scores four dimensions of each sentence, structural complexity, lexical sophistication, writing clarity and text coherence, and reports a grade with the hard sentences marked. The agreement section gives the correlation and the typical gap between the two on the same pages.

Q: Can I reuse the data?

Yes. The tables, charts and the per-page CSV are CC BY 4.0; link to this page. The crawl fetched one or two public pages per site with a named research bot that honours robots.txt, and the cached pages are not redistributed.

01

Key findings

What the crawl says.

Each finding is a complete, citable sentence with its number. Medians are per page; intervals are bootstrap 95% intervals of the mean.

The median public web page in this crawl reads at grade 8.0 on Orwellix's four-dimensional analysis and grade 10.3 on Flesch-Kincaid, across 1,530 pages from 7 kinds of site.
Median sentence 14.9 words; median Flesch Reading Ease 48.5; median 5.8% of sentences with a passive.
On the engine the seven slices sit within a narrow band: help centers easiest at a median grade of 7.8, universities hardest at 8.3. Flesch-Kincaid spreads them from 8.3 to 13.3.
Flesch-Kincaid ranks the slices SaaS pricing < SaaS homepages < News < Help centers < Top 1,000 homepages < Universities < Government; the engine ranks Help centers < Top 1,000 homepages < News < SaaS homepages < SaaS pricing < Government < Universities. Where they differ, sentence length is doing the work in the formula.
1% of pages score at grade 12 or above, the level the Nielsen Norman Group calls too much effort even for college-educated readers; 47% score below grade 8, its target for a broad consumer audience.[12][14]
Orwellix grades. On Flesch-Kincaid the figures are 29% and 20%.
Federal government pages run at a median Orwellix grade of 8.3 against 7.9 for the top-1,000 homepages, with sentences of 18.3 and 15.1 words.
Brown University's 2003 survey put state and federal sites at the eleventh grade on Flesch-Kincaid; this crawl's Flesch-Kincaid median for federal pages is 13.3.
SaaS homepages and their pricing pages read alike: median Orwellix grades of 7.9 and 7.9, sentences of 13.4 and 13.3 words, the shortest of any slice. Pricing pages carry more passives, 8.3% of sentences against 4.4%.
439 homepages and 190 pricing pages from the same Y Combinator companies. On Flesch-Kincaid pricing pages are the easiest slice at 8.3.
Help-center articles, the pages people read when something has gone wrong, score a median Orwellix grade of 7.8, with 6.7% of sentences in the passive.
65 help centers found at help. or support. subdomains of the SaaS slice.
University homepages score a median Orwellix grade of 8.3 and the longest words of any slice: 32.9% of words have seven letters or more.
348 four-year public and private nonprofit institutions from the federal IPEDS directory.
Flesch-Kincaid and Orwellix agree on the ranking of pages (correlation 0.45) but not on the level: the engine sits -2.6 grades from the formula on average, because it reads vocabulary and clause structure that two averages cannot see.
992 pages scored both ways.
Only 27 of the 70 news homepages yielded 80 words of sentence-shaped prose; what survives averages 856 words a page and reads at a median Orwellix grade of 7.9.
Headlines and teasers are dropped unless they form sentences of eight words or more, so this is standfirst and summary prose, not the articles; 23 news sites returned no HTML to a research crawler at all.

02

By slice

Seven kinds of page, side by side.

Top bar: median Orwellix grade. Bottom bar: median Flesch-Kincaid grade on the same pages. The dashed lines are the Nielsen Norman Group's consumer target and expert ceiling.

Reading level by kind of site, 2026

7 slices, 1,530 pages, crawled 23 Oct 2026

Orwellix grade is the engine's result on the first 1,200 words of extracted prose; Flesch-Kincaid uses the 1975 Navy coefficients on all extracted prose. Pages with fewer than 80 words of sentence-shaped text are excluded.

03

Slice table

Every slice, every measure.

Open a row for the quartiles, intervals, the share of pages under grade 8 and at 12 or above, and how many URLs the crawl attempted, fetched and kept.

Readability by slice, 2026

Help centers help. or support. subdomains of the same companies	65	7.8	9.3	55.6	14.4	6.7	2.4
Top 1,000 homepages Tranco list 56Z8N, ranks 1 to 1,000	258	7.9	10.6	47.1	15.1	4.7	4.9
News 70 widely read English-language news sites	27	7.9	9.3	55.6	16.1	6.8	5.9
SaaS homepages Y Combinator B2B companies, 600 sampled	439	7.9	9.1	53.5	13.4	4.4	3.0
SaaS pricing same companies, /pricing	190	7.9	8.3	57.9	13.3	8.3	3.7
Government GSA Site Scanning live federal sites, 600 sampled	203	8.3	13.3	32.6	18.3	10.5	12.8
Universities IPEDS 4-year institutions, 600 sampled	348	8.3	11.4	43.3	16.9	6.2	11.1

Showing 7 of 7 rowsHard sentences are those the engine flags as hard or very hard to read, as a share of the sentences it scored.

04

Most and least readable

The pages at each end.

Among pages with at least 150 words of prose. Sites are linked; the grade describes the public page on the crawl date, not the organisation.

Easiest to read

Easiest pages

trine.edu University homepages · Online Degree Programs with Flexibility and Support	5.6	11.6	20.2	182
www.lasalle.edu University homepages · Home - La Salle University	5.7	10.3	16.8	235
www.rrcc.edu University homepages · Red Rocks Community College	5.7	8.3	15.3	491
gbrain.io B2B SaaS homepages · GBrain · an agent that remembers for you	5.7	6.2	16.0	447
www.nfc.edu University homepages · North Florida College	5.8	8.1	12.4	174
www.lamar.edu University homepages · Lamar University is a Nationally-Ranked College in Texas - Lamar Unive	5.8	10.1	14.0	168
southbend.iu.edu University homepages · Indiana University South Bend	5.8	6.4	16.3	359
www.blacksmith.sh Help centers · Blacksmith Support \| GitHub Actions Help & Documentation	5.8	6.2	11.4	171
www.kaspersky.com Top 1,000 websites (homepages) · Kaspersky Cyber Security Solutions for Home and Business \| Kaspersky	5.9	9.0	14.2	598
www.one.com Top 1,000 websites (homepages) · Domains, Hosting, Websites & Email in One Place \| one.com UK	5.9	6.4	10.9	295
www.nbcnews.com News homepages · NBC News - Breaking Headlines and Video Reports on World, U.S. and Loc	5.9	8.4	13.2	658
www.census.gov US federal government sites · Developers	5.9	9.4	15.8	205
www.radford.edu University homepages · Radford University \| Find Your Place. Here.	5.9	9.9	16.5	181
sintem.co B2B SaaS homepages · Sintem	5.9	3.6	7.9	165
autosana.ai B2B SaaS homepages · Autosana \| AI Testing to Ship Faster	5.9	5.6	13.4	336

Showing 15 of 15 rows

Hardest to read

Hardest pages

www.councils.gov US federal government sites · About CDOC - Councils.gov \| Federal Executive Councils	14.6	19.5	29.9	986
wareiq.com Help centers · Helping Indian e-Commerce brands offer next-day delivery just like Ama	14.1	13.7	25.4	1,043
www.dhs.gov US federal government sites · Biometrics \| Homeland Security	13.8	17.9	25.3	835
www.washington300.gov US federal government sites · Washington 300	13.6	13.6	22.1	3,374
arkinlab.bio US federal government sites · Arkin Laboratory – Systems and Synthetic Biology	13.6	19.6	27.5	826
www.newhaven.edu University homepages · Home - University of New Haven	13.4	11.2	18.8	1,447
playbook.usds.gov US federal government sites · The Digital Services Playbook — from the U.S. DOGE Service	13.0	14.2	23.2	1,113
www.tamu.edu University homepages · Texas A&M University \| Texas A&M University	13.0	12.3	18.3	1,099
dra.gov US federal government sites · Homepage - Delta Regional Authority	12.0	15.9	21.2	892
www.dailymail.com Top 1,000 websites (homepages) · UK Home \| Daily Mail Online	11.9	14.4	31.0	8,854
www.dailymail.com News homepages · UK Home \| Daily Mail Online	11.9	14.4	31.0	8,854
www.stvincent.edu University homepages · This is Saint Vincent \| Latrobe, PA \| Saint Vincent College	11.9	12.9	21.5	1,654
www.whitman.edu University homepages · Whitman College	11.6	13.8	21.9	2,254
www.maritime.edu University homepages · Homepage \| Massachusetts Maritime Academy	11.6	15.5	19.7	1,141
unpkg.com Top 1,000 websites (homepages) · UNPKG	11.5	9.0	16.6	1,078

Showing 15 of 15 rows

Every page's scores are in the per-page CSV under Method and data, with the slice, host, word count and both grades, so any site can be looked up.

05

Distribution

Where the pages pile up.

Number of pages at each Orwellix grade. The shape, not the median, is what to compare with next year.

Pages by Orwellix grade

992 pages scored by the engine

Grades below 6 and 14 and above are pooled. Half-grades are rounded down.

06

Formula versus engine

Why a homepage can be 'grade 7' and still hard to read.

Flesch-Kincaid sees sentence length and syllables. Short marketing sentences full of product nouns look easy to it. The engine scores the structure, the vocabulary, the clarity and the coherence of each sentence.

0.45correlation of Flesch-Kincaid and Orwellix grades across pages
-2.6grades: the engine's average distance from the formula
29%of pages at grade 12+ on Flesch-Kincaid
1%of pages at grade 12+ on Orwellix

The classic formulas were fitted to schoolbooks and Navy manuals, prose written in full sentences[6][7]. Web copy is not: homepages run on fragments, pricing pages on lists, help centers on steps. The extractor keeps only sentence-shaped paragraphs, which is why word counts per page are modest, and the two measures still disagree in a consistent direction. One vendor study notes that the formulas' constants effectively cap scores near grade 12[11], and the million-policy study warns that they ignore organisation, formatting and semantic difficulty[21]. The engine's grade comes with the hard sentences marked, so a page can be fixed sentence by sentence rather than shortened blindly. For the formulas themselves see the Readability Score Reference.

07

Earlier studies

What was measured before, and how this differs.

Every published figure we could open, with the corpus it describes. Most measured pages that rank in search results; none measured whole slices of the web with a published method and dataset, which is the gap this study fills.

Prior measurements of web readability
		Measure	Published value	Source
Top-20 results for 1,000 keywords, 13,582 pages	2026	Mean grade, eight formulas	'12 or higher in nearly every case'	[11]
Privacy policies, 56,416 texts (2021)	2022	Flesch Reading Ease, mean and median; words per sentence	31; 31.8; about 27	[22]
COVID-19 news articles, 1,994	2022	Flesch-Kincaid grade, mean; sentence length	12.7 reliable, 14.4 unreliable; 23.7 and 26.4 words	[25]
Top-30 Google results for 30,000 queries	2021	Flesch Reading Ease, mean by position; grade	51.8 to 53.1; '11th-grade reading level'	[8]
Top-10 results for 15,000 keywords (food, marketing, engineering)	2021	Flesch Reading Ease, mean	food 69.7; marketing 60.2; engineering 49.6	[9]
Privacy policies, 1.07 million (2019 cohort)	2021	Flesch-Kincaid grade, median; length	13.2; 1,522 words	[21]
CDC COVID-19 pages, 68	2020	Flesch-Kincaid grade, median; words per sentence, median	11.0; 15.6	[19]
Top-20 Google results, 278,081 pages	2019	SMOG, mean and percentiles	13.06 years; 25th percentile 11, median 13, 75th percentile 15	[10]
Sign-in-wrap contracts of the 500 most popular sites	2019	Years of education (Flesch-Kincaid)	'more than 14 years'	[23]
Online health information, 7,891 websites in 157 studies	2018	Mean grade across scales	grade 10 to 15	[24]
Finnish university history pages, 7	2012	Flesch Reading Ease	11.3 to 46.8	[26]
US state and federal government sites, 1,663	2003	Flesch-Kincaid grade	'eleventh-grade level'; 67% at twelfth grade	[18]

Showing 12 of 12 rowsVendor studies are marked. Figures are quoted as published; formulas and extraction methods differ between studies, so compare with care.

Two findings recur in the earlier work. Pages that rank in Google cluster around grade 11 to 13 on the classic formulas, with no correlation between position and reading level[8][9][10]. And the documents nobody reads, privacy policies and terms of service, are the hardest of all: a median Flesch-Kincaid grade of 13.2 across a million policies, rising a full grade over twenty years[21][22]. What none of them give is a per-slice benchmark for ordinary public pages with a reusable method, which is what the table above is for.

08

Method, limitations, data

How the crawl was done, and what it leaves out.

Method

Sample frames are public, machine-readable lists: the Tranco top sites list (ID 56Z8N, ranks 1 to 1,000)[1], the GSA Site Scanning inventory of live federal sites[3], the IPEDS directory of four-year institutions[4], an unofficial mirror of the Y Combinator directory used only to find B2B companies' own websites[2], and a hand-compiled list of 70 widely read English-language news sites. Government, university and SaaS slices were sampled with a seeded shuffle of 600 each; the news list is published in full.

A named research bot fetched one page per site (two for SaaS: homepage and /pricing; help centers at help. or support. subdomains), honouring robots.txt under RFC 9309 including Crawl-delay, with a contact address in its User-Agent and no retries on refusals[5][28].
Prose was extracted by dropping navigation, headers, footers, forms and scripts, preferring the main or article element, and keeping sentence-shaped paragraphs of at least eight words. Pages with fewer than 80 words of prose, or not in English, were excluded.
Six formulas were computed on all extracted prose with one syllable heuristic; the passive share uses the Passive Voice Index detector; the Orwellix grade is the engine's result on the first 1,200 words.

Limitations

One page per site on one day. A homepage changes daily and a redesign changes everything; the per-page CSV records the fetch time.
Extraction is heuristic. Cookie walls, login walls and script-rendered pages yield no prose and are excluded, which removes some of the largest sites. The crawl table in each slice row says how many.
The SaaS slice is Y Combinator companies, which skews young and American; the help-center slice is the subset of those with a conventional help subdomain.
News homepages contribute summaries and standfirsts, not articles. Article-level readability is a different study.
Engine grades rest on the first 1,200 words of a page. Classic formulas and the engine were not calibrated on web fragments; treat both as descriptions of the prose, not of the reader.
No e-commerce product-page slice this year: the largest marketplaces refuse automated access and no free, licensed list of stores exists. It is planned for the next edition.

Data and files

Tables, charts and the CSVs are CC BY 4.0: reuse them with a link to this page. The crawled pages themselves are not redistributed.

09

Sources

28 documents, each one opened.

Sample lists, crawl standards, formula papers and every earlier measurement in the comparison table.

Le Pochat, V., Van Goethem, T., Tajalizadehkhoob, S., Korczyński, M., & Joosen, W. (2019). TRANCO: A Research-Oriented Top Sites Ranking Hardened Against Manipulation. NDSS 2019; list 56Z8N downloaded 23 Oct 2026 from tranco-list.eu. SourceOpened 23 Oct 2026
yc-oss (2026). Unofficial Y Combinator company directory API (industries/b2b.json). GitHub Pages mirror of the YC directory, last updated 22 Oct 2026; used only to find company websites. SourceOpened 23 Oct 2026
US General Services Administration (2026). Site Scanning program, live federal website inventory (site-scanning-live-filtered-latest.csv). digital.gov; file dated 22 Oct 2026. SourceOpened 23 Oct 2026
National Center for Education Statistics (2025). IPEDS Institutional Characteristics directory file HD2024. nces.ed.gov; 6,072 institutions with web addresses. SourceOpened 23 Oct 2026
Koster, M., Illyes, G., Zeller, H., & Sassman, L. (2022). RFC 9309: Robots Exclusion Protocol. IETF, Standards Track. SourceOpened 23 Oct 2026
Kincaid, J. P., Fishburne, R. P., Rogers, R. L., & Chissom, B. S. (1975). Derivation of new readability formulas for Navy enlisted personnel. Research Branch Report 8-75. SourceOpened 23 Oct 2026
Flesch, R. (1949). The Art of Readable Writing. Harper & Brothers; scan at archive.org. SourceOpened 23 Oct 2026
McKnight, T. (Portent) (2021). Study: How Content Readability Affects SEO and Rankings. Portent blog, 17 Jun 2021 (vendor). SourceOpened 23 Oct 2026
Pecánek, M. (Ahrefs) (2021). Flesch Reading Ease: Does It Matter for SEO? (Data Study). Ahrefs blog, 6 Jul 2021 (vendor). SourceOpened 23 Oct 2026
Antunes, H., & Lopes, C. T. (2019). Readability of web content: An analysis by topic. CISTI 2019; 278,081 pages, SMOG. SourceOpened 23 Oct 2026
Gillham, J. (Originality.ai) (2026). Best Readability Score To Rank in Google?. Originality.ai blog (vendor). SourceOpened 23 Oct 2026
Nielsen, J. (2015). Legibility, Readability, and Comprehension: Making Users Read Your Words. Nielsen Norman Group, 15 Nov 2015. SourceOpened 23 Oct 2026
Nielsen, J. (2005). Lower-Literacy Users: Writing for a Broad Consumer Audience. Nielsen Norman Group, 13 Mar 2005. SourceOpened 23 Oct 2026
Loranger, H. (2017). Plain Language Is for Everyone, Even Experts. Nielsen Norman Group, 8 Oct 2017. SourceOpened 23 Oct 2026
Google (2025). Search Quality Rater Guidelines, General Guidelines (11 Sep 2025). PDF; no occurrence of 'readability' or 'reading level'. SourceOpened 23 Oct 2026
Google Search Central (2025). Creating helpful, reliable, people-first content. developers.google.com; no mention of reading level. SourceOpened 23 Oct 2026
Schwartz, B. (2010). Google Reading Level Search Filter. Search Engine Roundtable, 10 Dec 2010 (the filter was later removed). SourceOpened 23 Oct 2026
West, D. M. (Brown University) (2003). Fourth Annual 'E-Government' Survey. press release of 12 Sep 2003, copy on the W3C WAI mailing list; 1,663 state and federal sites. SourceOpened 23 Oct 2026
Mishra, V., & Dexter, J. P. (2020). Comparison of Readability of Official Public Health Information About COVID-19 on Websites of International Agencies and the Governments of 15 Countries. JAMA Network Open 3(8):e2018033. SourceOpened 23 Oct 2026
Center for Plain Language (2024). 2024 Federal Plain Language Report Card. released 16 Dec 2024; judge-graded, not formula-based. SourceOpened 23 Oct 2026
Amos, R., Acar, G., Lucherini, E., Kshirsagar, M., Narayanan, A., & Mayer, J. (2021). Privacy Policies over Time: Curation and Analysis of a Million-Document Dataset. WWW '21; arXiv:2008.09159. SourceOpened 23 Oct 2026
Wagner, I. (2022). Privacy Policies Across the Ages: Content and Readability of Privacy Policies 1996–2021. arXiv:2201.08739. SourceOpened 23 Oct 2026
Becher, S. (2019). Are Online Agreements Readable?. The Regulatory Review, 4 Mar 2019 (author's essay on Benoliel & Becher, Boston College Law Review 60). SourceOpened 23 Oct 2026
Daraz, L., et al. (2018). Readability of Online Health Information: A Meta-Narrative Systematic Review. American Journal of Medical Quality 33(5):487-492; 157 studies, 7,891 websites. SourceOpened 23 Oct 2026
Zhan, K., Li, Y., Osmani, R., Wang, X., & Cao, B. (2022). Data Exploration and Classification of News Article Reliability: Deep Learning Study. JMIR Infodemiology 2(2):e38839; 1,994 COVID-19 news articles. SourceOpened 23 Oct 2026
Karhu, M., Hilera, J. R., Fernández, L., & Ríos, R. (2012). Accessibility and readability of university websites in Finland. Journal of Accessibility and Design for All 2(2):178-190. SourceOpened 23 Oct 2026
Harsel, L. (Semrush) (2026). How We Built a Content Optimization Tool for AI Search [Study]. Semrush blog, 14 Jan 2026 (vendor); 304,805 AI-cited URLs, no readability metric reported. SourceOpened 23 Oct 2026
Common Crawl (2026). Frequently asked questions (crawler behaviour). commoncrawl.org. SourceOpened 23 Oct 2026

10

Cite and track

Cite this page.

Cite this page

APA

Orwellix. (2026). The State of Web Readability 2026. Orwellix. https://orwellix.com/resources/state-of-web-readability

Plain

The State of Web Readability 2026 (Orwellix, 2026): https://orwellix.com/resources/state-of-web-readability

First published 25 Oct 2026. Reuse the tables and charts with attribution under CC BY 4.0.

Changelog

25 Oct 2026First edition: seven slices crawled on 23 Oct 2026 (Tranco top 1,000, 70 news sites, 600 federal government sites, 600 universities, 600 Y Combinator SaaS companies' homepages, pricing pages and help centers), scored with six formulas, the Passive Voice Index detector and the Orwellix engine.

11

Questions

The questions this page gets asked.

01What reading level is most web content written at?

In this crawl, the median page across all slices sits where the summary tiles say, with the top-1,000 homepages and SaaS pages at one end and government and university pages at the other. Prior studies of pages that rank in Google put the typical page around grade 11 to 13 on the classic formulas; this study reads whole sites' public pages rather than search results, scores them with the same extractor and the same formulas, and adds the Orwellix four-dimensional grade with the hard sentences counted.

02What grade level should my website be?

The Nielsen Norman Group's advice, which most style guides repeat, is 8th grade for a broad consumer audience, 6th grade on landing pages for lower-literacy readers, and 10th to 12th for expert readers. Google states no target and its rater guidelines do not mention reading level. Use the slice closest to your audience in the table as a benchmark, and the hard-sentence share rather than the grade alone.

03Are websites getting harder or easier to read?

This is the first edition, so the trend will come from next year's crawl of the same slices with the same method. The only long-run measurements we could open are for privacy policies, which got longer and about a grade harder between 2000 and 2019, and for government sites in 2003, which read at the 11th grade on Flesch-Kincaid.

04Why score homepages? They are mostly headlines.

Because that is what people land on. The extractor keeps only sentence-shaped paragraphs of at least eight words, drops menus, cookie banners and footers, and discards any page with fewer than 80 words of prose, so the score describes the sentences a visitor actually reads. The word count per page is in the table, and the per-page CSV lets you filter to longer pages.

05How does the Orwellix grade differ from Flesch-Kincaid here?

Flesch-Kincaid counts words per sentence and syllables per word. Homepage copy is short-sentenced and is read as easy by the formula even when it is dense with product jargon. Orwellix scores four dimensions of each sentence, structural complexity, lexical sophistication, writing clarity and text coherence, and reports a grade with the hard sentences marked. The agreement section gives the correlation and the typical gap between the two on the same pages.

06Can I reuse the data?

Yes. The tables, charts and the per-page CSV are CC BY 4.0; link to this page. The crawl fetched one or two public pages per site with a named research bot that honours robots.txt, and the cached pages are not redistributed.

See where your own page lands

Paste your homepage copy into the free Orwellix readability checker: one grade on the same four dimensions used here, with the hard sentences marked. Ten checks a day, no account.

Open the free checker

The Content Brief + QA Scorecard

Brief the writer, score the draft, generate the feedback

Get the download

All resources · The Readability Score Reference · The Passive Voice Index

The public web reads at grade 8.0. 1% of pages sit at grade 12 or above.