Text forensics · search quality
Know your content.
Understand its signals.
Analyze writing patterns, examine possible AI-generation signals, and audit web pages for technical SEO issues — all in one transparent workspace.
No sign-up. No stored text. Results are signals, not verdicts.
Features
One workspace, honest outputs
AI content analysis
A pluggable detector interface that reports raw scores, thresholds and uncertainty from a real model — or says plainly that no model is configured.
Paragraph-level findings
Every paragraph is shown with its own model score when the model supports it, plus locally measured writing signals clearly labeled as heuristics.
Experimental attribution
Model-family attribution is kept separate from detection and stays disabled until a classifier for it has been evaluated.
Technical SEO audit
Status codes, redirects, robots rules, canonicals, noindex, headings, alt text, JSON-LD syntax and a sample of internal links — from the live response.
Downloadable reports
Export JSON, print or save as PDF, and copy an executive summary. Every number comes from the analysis that ran.
Privacy-first
Text is processed in memory and not stored. Nothing is sent to a third-party AI service without your explicit consent.
How it works
What actually happens to your text
- 1
Validate
Length, character and language limits are checked. Your text is not rewritten.
- 2
Segment
Paragraphs and sentences are split from the original text, keeping their exact wording.
- 3
Analyze
Writing signals and content-quality rules run locally. If a detector is configured (and you consent), it scores the text.
- 4
Report
Summaries are built only from the outputs that exist. Missing checks are listed as unavailable.
Methodology & limitations
Probabilistic, not proof
AI-text detectors estimate how closely text resembles what a model was trained to recognise. Scores shift with topic, length, editing and writing style, and human writing can be incorrectly flagged — non-native and highly formal writing in particular. TextScope reports raw scores as raw scores, marks results near the threshold as inconclusive, and never treats “human-like” as proof of human authorship.
Read the methodology- Never invented: detection scores, attribution percentages, index status or crawl results.
- Never claimed: Google’s private ranking or AI-detection behaviour.
- Always shown: which checks ran, which did not, and where each finding came from.
FAQ
Questions
How accurate is the AI detection?
It depends entirely on the configured model and the kind of text. TextScope publishes no accuracy figure it has not measured. All detectors make mistakes: human writing — especially formal, edited or non-native writing — can be flagged, and generated text can pass. Treat results as signals, never as proof.
Is my text stored or shared?
No. Text is analyzed in memory for one request and discarded. If the operator connects a self-hosted inference service, your text is sent to it only after you tick the consent box.
Which formats are supported?
Plain text pasted or typed into the editor, in English, between 50 words and 20,000 characters. File uploads (PDF, DOCX) are not supported yet.
Can TextScope tell me if Google indexed my page?
No. The URL audit shows whether a page is crawlable and indexable from its public response. Actual index status requires Google Search Console access for a property you own, which is not connected in this deployment.
Can it tell which AI wrote the text?
Not today. Attribution to a provider or model family needs its own evaluated classifier; until one exists the feature shows as unavailable instead of guessing.
Is it free?
Yes, and it runs on free hosting. That also means usage is rate-limited and not unlimited.