A grammar mistake in business copy is a language error that reaches the reader: its where it should be it's, a verb that does not agree with its subject, a word spelt the way it sounds. It is not a Google ranking factor. It is, however, one of the few things a reader notices without looking for it.
For this article I measured how many spelling errors appear in the pages currently on the first page of Google UK, across three unrelated sectors: 17 pages, 51,533 words. I found no unambiguous misspellings at all, but 7 of the 17 pages break British spelling conventions, and they rank on page one anyway.
What Google says about grammar mistakes
Google has made three public statements on the subject, collected by Search Engine Journal in its ranking factors guide. In 2011 Matt Cutts said spelling and grammar were not direct ranking signals, but that in tests better-written sites came across as more reliable. In 2017 John Mueller explained that grammar does not affect search engines, it affects users. In 2021 Mueller called it a grey area: if a text is so badly written that the topic becomes unreadable, the problem stops being cosmetic and becomes one of comprehension.
The official documentation is more useful than the statements. In Creating helpful, reliable, people-first content, Google lists the questions a publisher should ask about their own content. Two concern the care taken over the text: Does the content have any spelling or stylistic issues? and Is the content produced well, or does it appear sloppy or hastily produced? These are self-assessment questions, not algorithmic criteria. Google puts them in the same list where it asks whether the content provides original information and whether it is written by someone who knows the subject.
In short: spelling is not an item in the algorithm in the way links or speed are. If you want the map of what does carry weight, we have put it together in our guide to ranking factors.
The test: 17 first-page pages on Google UK, 51,533 words
I ran the test in September 2026 on three queries from sectors far apart: best running shoes (e-commerce and reviews), seo agency (B2B services) and how to write a blog post (editorial content). For each I took the organic results on page one of Google UK (DataForSEO, Google UK, September 2026), extracted the editorial text of the pages (paragraphs, lists, headings and table cells, leaving out scripts, stylesheets, menus and footers) and searched for twenty forms that are wrong in English with no room for interpretation: should of, could of and would of, alot, definately, seperate, recieve, occured, untill, accomodate, thier, begining, neccessary, wierd, independant, succesful, it's own and there own, it's in front of a noun (it's website, it's customers), your not, their is, more then and better then, and loose in place of lose. I also counted American and British spellings of the same word families (optimise/optimize, organise/organize, colour/color, favourite/favorite, behaviour/behavior and a few others).
The three SERPs returned 29 organic results. Twelve dropped out of the sample: two are YouTube videos, two did not respond to the request, six blocked it (three Reddit threads, Quora, Medium and one agency site) and two load their text via JavaScript and expose none in the HTML. That leaves 17 pages and 51,533 words. This is the result.
| Measure | Value |
|---|---|
| Pages analysed | 17 |
| Total words | 51,533 |
| Unambiguous misspellings found | 0 |
| Pages written entirely in American spelling | 5 of 17 |
| Pages mixing British and American spelling | 2 of 17 (behaviour and organize on the same page, optimisation and organize on another) |
| Pages that break British conventions | 7 of 17 (41%) |
Not one of the twenty classic misspellings appears in more than fifty thousand words. What does appear is spelling that a British reader recognises as foreign: Wix, HubSpot, LinkedIn and two American running sites rank in the UK with optimize, color and favorite, and one of them sits in the AI Overview as the main source. Two pages, one of them from a UK agency selling SEO, switch between the two systems within the same text. None of them has lost page one over it.
We ran the same test on Google Italy with eleven common Italian misspellings: 6 of 25 first-page pages contained at least one, including agencies that sell web copywriting, and they ranked too. Either way, the ranking does not move.
The limits of the test need stating. The sample is small and is not statistically representative of the UK web. I only counted forms that are 100% wrong, so agreement errors, punctuation and syntax are left out, and those are far more common and far harder to detect automatically. A result of zero means the most obvious errors are rare on page one, not that the pages are error-free. Anyone who wants to repeat the measurement on their own queries has the full method above: twenty regular expressions and the text of the pages.
What a mistake really costs
If the position does not move, something else does. The most cited study on the point is by Julie Boland and Robin Queen of the University of Michigan, published in PLOS ONE in 2016: 83 participants read 12 email replies to an advert for a housemate and rated the writers on a scale of 1 to 7. Writers whose messages had typos lost 0.47 points on average, those with grammatical errors 0.22. The sample was American and the texts were emails, not web pages, so the number does not transfer directly to a company website. What does transfer is the mechanism: the reader's judgement of the writer changes before the reader notices it has changed.
In B2B this matters more than elsewhere, because the text is not selling a £30 product. On an industrial project we manage, the Macropix case, the main keyword gets 4,400 searches a month in Italy and the page that captures it brings in quote requests for custom installations. Over seven years that project has collected around seven thousand contact requests. The reader who lands there is a procurement manager deciding whether to write to a supplier they do not know, and the only thing in front of them to judge by is the page. The same reasoning applies to every service page: care over the text is one of the few signals of experience and trustworthiness that comes across in three seconds of reading, and it acts on the conversion rate, not on the position.
The five mistakes you actually find in business copy
The classic misspellings did not turn up in the test because spellcheckers have been catching them for decades. The mistakes that survive are the ones where the wrong word is a real word, plus one that is specific to British English.
| Mistake | Correct form | Does an automatic checker catch it? | Does it affect position? | Does the reader notice? |
|---|---|---|---|---|
| it's website, it's customers | its website, its customers | Sometimes: both are real words | No | A lot: it is the signature mistake |
| your welcome, your not | you're welcome, you're not | Sometimes | No | Often |
| there, their, they're | depends on the sentence | Sometimes | No | People who write for a living, yes |
| more then, better then | more than, better than | Rarely | No | Little, but it signals carelessness |
| optimise and optimize on the same page | one system, applied throughout | Only if the checker is set to the right variety of English | No | British readers do |
The last one deserves a line of its own, because it is not strictly an error. The -ize ending is not an Americanism: it is the house style of Oxford University Press, known as Oxford spelling, and it is correct British English. A British page written consistently with optimize and colour is fine. What reads as a mistake is inconsistency: optimise in the heading and optimize in the paragraph below, or color next to behaviour. That is what the two mixed pages in the test do, and it is almost always the trace of copy pasted from different sources or tools.
In 2026 the mistake that costs most is not grammatical
There is a practical reason why hunting for typos has become less urgent: typos are disappearing on their own. A text generated or rewritten with ChatGPT, Gemini or Claude has flawless syntax and does not misspell a word. It gets other things wrong, and no spellchecker sees what it gets wrong.
The mistakes we find when we review AI-assisted copy are always the same four: product names mangled or altered, plausible but invented figures, dates and regulations cited in the wrong place, and American defaults in British copy (spelling, but also terms such as zip code or sales tax on a page for UK buyers). Each of these passes straight through LanguageTool and through Word's checker, because they are grammatically correct sentences that say something false or something written for another market.
For Google the difference is clear. The spam policies do not mention spelling anywhere, while they devote a whole section to scaled content abuse, defined as producing many pages for the primary purpose of manipulating search rankings and not helping users, and they explicitly cite the use of generative tools to create pages without adding value. The system does not penalise you for an apostrophe. It ignores you for a page that says nothing. We have gathered the data on how Google really treats these texts in the article on AI content and SEO, and the policy side in AI-generated content.
There is also an effect on answer engines. When ChatGPT or Perplexity put together the profile of your company, they build it from how your name, your products and your figures appear across pages. If the product name is written three different ways on three pages of your site, the entity does not consolidate and the generated answer stays vague. That is a consistency problem, not a grammar problem, and it is what counts for appearing in ChatGPT answers and more generally for generative engine optimisation.
How to check a site without rereading all of it
The check we run before publishing takes three steps and twenty minutes per page.
- The automatic scan, once for the whole site. Screaming Frog has a built-in spelling and grammar check that works on the entire crawl, and you can set the language to English (UK); alternatively the LanguageTool API returns the same errors in a format a script can read. It clears out the typos, which are the easy part.
- Fact-checking, by hand, only on the things that can be wrong. Proper names, figures, percentages, dates, prices, units of measurement, names of regulations and products. If a number cannot be found in a primary source, it comes out of the text. This is the step no tool does for you, and it is the one that contains the expensive mistakes.
- A read by someone who did not write the text. Writers reread what they meant to write, not what they wrote. One different person and a single read is enough.
The order in which to check pages is not the order of the site: service pages and the pages that receive contact forms first, then product pages, then the blog. The full method for writing copy that passes both checks, language and facts, is in the guide on how to create SEO content, and the blog side is in blog SEO. If the site is large and the review needs to be set up as a process rather than an emergency, it is one of the jobs we do within our SEO projects; the results of some of them are in projects.
The rule I took from the test is this: check spelling on the pages that convert, check facts everywhere. Getting the two the wrong way round is the most common way to spend a day removing apostrophes while a wrong price stays on the page for months, and unlike the apostrophe, that one really does lose you a customer.
Frequently asked questions
No. Google does not use spelling and grammar as a direct ranking factor: Matt Cutts said so in 2011 and John Mueller in 2017 and 2021. In our September 2026 analysis of 17 first-page pages on Google UK, 7 broke British spelling conventions and still ranked on page one; on Google Italy, 6 of 25 first-page pages contained at least one clear misspelling.
There is no penalty for spelling and Google's spam policies do not mention it. Mistakes become a ranking problem only when they make the text incomprehensible, because at that point the search engine cannot tell what the page is about. Google's documentation asks publishers whether their content has spelling or stylistic issues, but as a self-assessment question about quality.
In a sample of 17 pages and 51,533 words taken from page one of Google UK for three queries in different sectors, we found none of twenty classic misspellings, such as should of, alot or recieve. We did find 5 pages written entirely in American spelling and 2 that mix British and American forms. The count leaves out agreement, punctuation and syntax, so the real number of errors is higher than zero.
Yes, and it costs little: Screaming Frog includes a spelling check across the whole crawl and the LanguageTool API does the same from a script. It is not enough on its own: a checker does not see wrong product names, invented figures and misquoted dates, which are the mistakes that lose you a customer. Those are checked by hand on the pages that receive contact requests.
A wrong figure. A typo affects how the reader perceives you: in the study by Julie Boland and Robin Queen published in PLOS ONE in 2016, writers with typos lost 0.47 points on average on a scale of 1 to 7. A wrong number or price, on the other hand, goes all the way to the negotiation and stays there.
