Small SEO Tools AI detector: how accurate is it?
Short answer We found no published independent test of it. Its page claims 99.9% accuracy but shows no test, and the same site sells a humanizer it says can bypass all AI detectors. Independent studies of similar tools found many flag human writing as AI. Treat its score as a hint, never as proof.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callHow accurate is the Small SEO Tools AI detector?
No one outside the company has shown it, as far as we found. The tool's page says "Accuracy 99.9%" and "The results of our AI checker are fully accurate" (SmallSEOTools: AI Content Detector). It doesn't say what texts it was tested on, how many, or how often it calls human writing AI. We found no peer-reviewed study that tested it.
Its own FAQ, under the heading "How Accurate is the SmallSEOTools AI Writing Detector?", gives no number at all. It only says the tool will help you tell the two apart.
| What the page says | What it doesn't say |
|---|---|
| 99.9% accuracy | On what texts, and who tested |
| Shows a percentage of AI vs human text | How often human text gets an AI score (the false-positive rate) |
| Works in French, German, Spanish and more | Whether accuracy holds outside English |
| Free, no signup, unlimited checks | Which AI models it was trained to catch |
What does the free version do?
You paste text or upload a file, press the button, and get two numbers: AI-written and human-written share. It guesses from patterns in the words; it can't know who actually wrote them. The limits, from the tool's page on November 16, 2026:
- 2,000 words per check. The page says up to 2,000 words are used.
- Unlimited checks, no account, no fee.
- Your text isn't stored, the page says: entered data "is automatically removed" after the check.
The same site also has an AI Humanizer, a tool that rewrites text to read as human. Its page says it can "modify the text to bypass all AI detectors" (SmallSEOTools: AI Humanizer). Both claims can't be fully true at once: if a rewrite beats every detector, it beats this one too.
Why should I doubt a 99.9% claim?
Because detector makers have claimed high numbers before, and tests didn't hold up.
- A regulator acted on one. The FTC says Workado promoted its AI Content Detector as "98 percent" accurate, "but independent testing showed the accuracy rate on general-purpose content was just 53 percent" (FTC, April 28, 2025). Its order requires proof before such claims.
- OpenAI withdrew its own detector. In OpenAI's tests it caught 26% of AI text and labeled human text as AI 9% of the time. On July 20, 2023, OpenAI said it was "no longer available due to its low rate of accuracy" (OpenAI).
- Edited AI text slips through. OpenAI also warned: "AI-written text can be edited to evade the classifier" (OpenAI). So an AI draft you edit by hand can score human: a low score proves nothing either.
How often do detectors flag human writing?
Often enough to matter. This is the false positive: real human writing scored as AI.
- Non-native writers get flagged. Stanford researchers ran seven detectors on 91 essays by non-native English writers. On average 61.22% were called AI-generated, and 89 of 91 were flagged by at least one tool (Liang et al., 2023).
- Translated text gets flagged more. A test of 14 detectors found six produced false positives, "with the risk increasing dramatically for machine-translated texts" (Weber-Wulff et al., 2023). If you write in Spanish, Chinese or French and translate to English, you may see more AI scores.
- Short text is shaky. OpenAI said its detector was "very unreliable on short texts (below 1,000 characters)" (OpenAI). That's about 150–200 words: a service blurb, a product description, an About page.
Neither study tested Small SEO Tools. They show how this kind of tool fails, not its exact rate.
How do I test it on my own writing?
Ten minutes tells you more than the 99.9% claim. Use texts where you know the answer.
- Old human text. Paste 300–500 words you or your staff wrote years ago, before AI chat tools: an old newsletter, a proposal, an email. Any AI score here is a false positive.
- Pure AI text. Ask a chatbot for 300 words on your service. Paste it unedited. A low AI score here is a miss.
- Your real process. Take that AI draft, add your own prices, steps and job details, and paste again. This is closest to what you'd publish.
- Short text. Paste a 100-word service blurb you wrote yourself.
- Repeat one. Run test 1 again and compare the numbers.
Write the five scores down. If test 1 or 4 comes back mostly "AI," you've measured its false-positive risk on your own writing. Don't accuse anyone, or rewrite anything, on its score.
Does a Small SEO Tools score matter for Google?
No. Google judges the page, not how it was written:
"Our focus on the quality of content, rather than how content is produced, is a useful guide that has helped us deliver reliable, high quality results to users for years." (Google Search Central Blog, February 8, 2023)
Google has never said it uses a third-party detector score. What to do instead is in AI detectors and SEO, and the policy itself in Does Google penalize AI content?
Example
A bakery checking a freelancer's blog posts
The owner pastes a freelancer's post; it scores 70% AI. Before saying anything, she runs one of her own 2019 newsletters, which scores 40% AI. The score can't settle it, so she checks the post itself: two prices are wrong. She asks for those fixes, not a "human" rewrite.
A contractor who writes in Korean first
He drafts in Korean, translates with a free tool, then edits. His pages score high for AI. That's the pattern the 14-detector study found for translated text. He leaves the wording alone and adds job photos and local permit steps.
An accountant over the word limit
Her 3,500-word tax guide is over the 2,000-word limit, so only part gets checked. She'd have to split it in two to see the rest, and each half is a separate guess.
FAQ
Is the Small SEO Tools AI detector free?
Yes. Its page says no signup, no fee and unlimited checks, up to 2,000 words each.
Does it work on languages other than English?
The page says it supports French, German, Portuguese, Spanish, Italian and more. It publishes no accuracy figure for any language.
Can a score from it prove someone used AI?
No. It gives no false-positive rate, and studies of similar tools found high ones. OpenAI said its own detector should not be "a primary decision-making tool."
Does it keep the text I paste?
Its page says entered data is removed from its databases once the check is done. Still, don't paste client records or anything confidential into any free online tool.
Different for you?
- Worried a detector score hurts rankings: AI detectors and SEO
- Want Google's actual rule on AI pages: Does Google penalize AI content?
- Deciding which writing jobs to give AI: AI content and SEO
- Want pages written and checked for you: SEO agent