FAQs

You can upload html, htm, txt, rtf, xml, md, pdf, docx, doc, pptx, ppt, odt, epub, odp, ppsx, pages, xlsx and xls files.

Yes. You can upload and scan up to five files at once.

Your document is not added to our shared comparison database unless you choose that option when setting up your scan. If you opt in, your document is added to the internal database so that it can be compared with future submissions, while your own scan gains access to more than one million previously submitted documents. You can ask us to remove your document afterwards at no charge. If you do not opt in, your document will not be added to the shared comparison database. Your uploaded documents and scan results are not published publicly.

We do not notify your university, school, employer or tutor when you scan a document. However, if you choose to add your document to the shared comparison database, the same content could later appear as a match if it is checked against that database again. If you do not want this to happen, simply leave the internal database option switched off (it is switched off as default).

Most scans are completed within a few minutes, although larger documents and more detailed scan options can take longer. Once your scan has been submitted, you can leave the processing page if you wish. Your completed report will appear under My scans when it is ready.

Our AI detector achieves over 99% accuracy while maintaining an exceptionally low false-positive rate. In large-scale testing of the current detection model, the standard balanced setting correctly identified 99.21% of AI-generated texts and 99.97% of human-written texts, with a false-positive rate of just 0.026% – approximately 0.03%. Accuracy can vary according to the type and length of text and how extensively AI-generated material has been edited, so results should always be considered in context.

PlagPointer lets you choose between three AI-detection sensitivity settings:
Extra Safe – the most cautious option. It is designed to minimise false positives and is best suited to identifying relatively unmodified AI-generated text.
Balanced – the default and recommended setting for most scans. It balances detection of edited AI text against a very low false-positive rate.
Extra Sensitive – the most aggressive option. It is designed to identify AI-generated text that has been heavily edited, rewritten or passed through a “humaniser” or text spinner, but this increased sensitivity also produces a higher false-positive rate.
Changing this setting affects AI detection only. It is separate from the Low, Medium and High plagiarism-sensitivity setting used to control how thoroughly PlagPointer searches for text matches.

The detector can identify content produced by leading generative AI systems including GPT-5, GPT-4/ChatGPT, Claude, Gemini, DeepSeek, Llama, Bloom, Rytr and Jasper, as well as other models. Detection is based on the underlying patterns used to generate AI-written text rather than relying only on a list of individual AI products. The technology is continually developed as new models and generation techniques appear.

Yes. The technology is designed to identify more than text copied directly from ChatGPT or another AI tool. It can detect AI-generated writing that has undergone human editing, rewriting or paraphrasing. The underlying detector has three sensitivity models, ranging from an Extra Safe setting designed to minimise false positives through to an Extra Sensitive model designed to identify AI text that has been heavily modified or passed through a “humaniser” or text spinner.

Yes. AI-generated content does not have to make up an entire document. The detector is designed to identify AI-written passages even when they have been mixed with genuine human writing.

Generative AI produces text using statistical patterns that differ in subtle ways from natural human writing. The detection technology behind PlagPointer has been trained using extremely large datasets of human-written and AI-generated material, allowing it to recognise these differences. Its analysis can go beyond a single overall score. It can examine individual passages and writing patterns, identify phrases that occur statistically more often in AI-generated writing, and compare text with AI-generated material already published elsewhere.

It can. No responsible AI detector should claim that false positives are impossible. The current balanced model produced a false-positive rate of just 0.026% in large-scale testing. An Extra Safe sensitivity model reduced this further to 0.009%, while the more aggressive Extra Sensitive model had a false-positive rate of 0.05%.

No AI detector should be treated as conclusive proof of authorship by itself. An AI detection result is an important indicator that can help identify material for further review. For academic or professional decisions, the result should be considered alongside the surrounding evidence, such as drafts, writing history, sources, institutional rules and an explanation from the writer.

AI detection is available in 30+ languages, including English, French, Spanish, German, Italian, Portuguese, Dutch, Chinese, Japanese, Arabic, Hindi, Korean, Polish, Russian and many more.

AI detection requires at least 350 words in each file. If a file is shorter than this, it can still be checked for plagiarism, but the scan will not return an AI-detection result for that file. AI detection is also not available for source-code files.

The checker looks beyond simple copy-and-paste plagiarism. It distinguishes between:
Identical matches – text copied word for word.
Minor changes – copied material where small alterations have been made, such as changing a tense or substituting a few words.
Paraphrased content – substantially rewritten text that still conveys the same underlying meaning as the source.
This allows the checker to identify attempts to disguise copied material rather than simply looking for identical sentences.

Yes. Advanced semantic analysis allows the checker to identify text that has been substantially rewritten while retaining the meaning or ideas of the original source. It is therefore capable of finding plagiarism that a simple word-for-word comparison could miss.

Yes. Optional cross-language plagiarism checking can identify content that has been copied from a source in another language and then translated. This is particularly useful where conventional same-language plagiarism checks would fail because the wording has changed completely during translation. Cross-language checking requires additional processing, so additional credits are used when you select it.

Standard plagiarism detection supports more than 100 languages, including English, Spanish, French, Portuguese, Arabic, Russian, German, Greek, Chinese and Japanese.

Yes. You can exclude references, quotations, citations and your table of contents from the plagiarism analysis. This helps prevent legitimate quoted or referenced material from artificially inflating the similarity percentage.

Plagiarism sensitivity controls how extensively the checker searches for potential matches. A high-sensitivity scan looks for more potential matches but can take longer to complete. A low-sensitivity scan is quicker and focuses on the strongest or most relevant matches. The standard setting provides a balance between the two. This setting relates specifically to the plagiarism scan and is separate from the sensitivity models used by the AI detector.

No. A similarity score tells you how much content matches material found elsewhere; it does not decide whether plagiarism has occurred. Properly quoted material, references, commonly used phrases and legitimately reused text can all create matches. The report should therefore be reviewed together with the identified sources and the context in which the matching material appears.

Yes. In Advanced scan settings you can enable cheat / text-manipulation detection. This can flag possible hidden characters, lookalike character substitutions and invisible text that may have been used to interfere with plagiarism checking. An alert indicates possible manipulation; it is not proof of why the text was formatted that way.

Yes. In Advanced scan settings, choose “Exclude specific websites” and enter the domains you do not want included in the results. You can also choose “Search only specific websites” if you want the scan restricted to particular domains. Leave the setting as “Search the whole web” for the normal scan.

You can choose to exclude references, quotations, citations and the table of contents. For HTML files you can also exclude titles/headings and repeated website header/footer content. For source-code files you can exclude code comments. These options only affect plagiarism matching; they do not rewrite or remove content from your original file.

Yes. PlagPointer handles supported source-code files for plagiarism checking, including common JavaScript, TypeScript, Python, Java, PHP, C/C++, C#, Go, Ruby, Swift, shell, Scala and CSS files. Source-code files are checked for plagiarism only: AI content detection is not supported for code. The maximum source-code file size is 3 MB per file.

Yes. PlagPointer is pay-as-you-go. You buy credits once, use them whenever you like, and they never expire. There is no monthly fee, no auto-renewal and nothing to cancel. Most subscription checkers reset your credit allowance at the end of each billing period and take back anything unused – whether you’re billed monthly or annually.

One credit covers up to 250 words and includes both plagiarism and AI detection. A document containing 1–250 words uses one credit, 251–500 words uses two credits, and so on. Optional cross-language checking uses additional credits. Before you pay for a scan, PlagPointer calculates the exact number of credits required so you can see the cost before proceeding.

No. Credits you purchase are stored in your account and do not expire, so you can buy a larger bundle for a lower price without having to use it within a particular period.

No. PlagPointer is pay-as-you-go. Purchase the number of credits you need and use them whenever you wish. Larger credit bundles reduce the cost per 250 words, with current prices starting from 18p per credit.

Payments are processed securely by Stripe. We accept Visa, Mastercard, American Express, Discover, Diners Club, JCB and China UnionPay.

If a bundle is entirely unused, contact us with 21 days for a full refund. If you have used part of a credit bundle with the 21 days, you may request a refund of the unused balance. We will recalculate the credits already used at the price that would have applied to the quantity actually used at the time of purchase and deduct that amount from the price originally paid. For example, if you purchase a 50-credit bundle but use 10 credits before requesting a refund, the 10 credits used may be recalculated at the applicable 10-credit bundle rate, with the remaining balance refunded. We do not offer voluntary cash refunds where the amount refundable under this policy would be less than £5.

A 2,500-word essay uses 10 credits – from £4.99. Credits are cheaper if you buy a larger bundle. A 10,000-word dissertation chapter uses 40 credits – so you would buy a bundle of at least 50 credits (£14.99) and have 10 credits leftover. One credit covers 250 words and includes both plagiarism and AI detection. You see the exact credit cost before anything is deducted.

Click Register in the menu, then choose a username and enter your email address and password.

Click Login, then select Lost your password? and enter the email address associated with your account. We’ll send you a password-reset link.