HomeAI Detector › Bulk

Bulk AI detector — 500 files at once via UI or REST API.

Single-paste workflows collapse when a teacher is staring at 200 essays a term, an agency editor has 40 freelance drafts on a Friday, or a content platform screens thousands of user submissions a day. TextSight bulk detection runs on two surfaces backed by the same classifier. The Bulk Scan UI accepts paste, drag-and-drop file upload across 11 formats (.docx, .pdf, .txt, .md, .rtf, .odt, .epub, .html, .pptx, .xlsx, .csv), and ZIP archives that unpack into one queue, up to 500 files per batch on Business. The REST batch endpoint at api.textsight.ai/scan/v1/batch takes an array of documents in one call and returns an array of scored results with webhook callbacks for asynchronous pipelines. Both ship on the Business tier with five workspace seats, shared scan history, and a workspace-wide audit log.

Start bulk scanning free See Business pricing
500 files per batch 11 file formats REST API + webhooks 5-seat workspace
TextSight pricing

Bulk lives on the Business tier.

Free, Starter, and Pro cover individual reviewers. Business adds the Bulk Scan UI at 500 files per batch, the REST batch endpoint, webhook callbacks, 5 workspace seats, the audit log, multi-client workspaces, and white-label PDF reports. Yearly billing saves 25%.

Free
$0/forever

 

Try the dashboard on a real document first.
  • 3 scans / day
  • 5,000 chars per scan
  • Sentence-level highlights
  • Dashboard UI only
Start free
Starter
$7.49/month

Billed $89.88/year, Save $30

Solo reviewer, dashboard plus extension.
  • 20 scans / day
  • 20,000 AI rewriter words/mo
  • Chrome extension
  • Email support
Get Starter
Pro
$14.99/month

Billed $179.88/year, Save $60

Unlimited scans, file upload. No batch API.
  • Unlimited scans
  • 50,000 AI rewriter words/mo
  • File and URL upload
  • Priority support
Get Pro

Yearly billing saves 25%. View full pricing →

Two surfaces, one classifier

Bulk UI for reviewers, REST API for pipelines.

Both paths read from the same workspace and run the same model. The UI is the right answer when a human is sorting and triaging; the REST endpoint is the right answer when another system is the trigger.

Path one: the Bulk Scan UI

Open the dashboard, click Bulk Scan, and paste, drag-and-drop, or upload a ZIP of up to 500 files. The unified extractor handles .docx, .pdf, .txt, .md, .rtf, .odt, .epub, .html, .pptx, .xlsx, and .csv. Each item lands in a queue row with Authenticity Score, band label, word count, and sentence highlights. Sort ascending by score to triage the highest-risk submissions first. Click any row to open the reading view with green, yellow, and red highlight spans for Human, Mixed, and AI sentences. The queue persists across page refreshes and shares with every seat on the workspace.

Path two: the REST batch endpoint

POST an array of documents to api.textsight.ai/scan/v1/batch with an Authorization Bearer header and get an array of scored results in one call. Each result carries the score, the band, per-sentence labels with start and end offsets, and per-model confidence. Webhook callbacks let a CMS, LMS, or moderation pipeline submit a batch and forget the connection. The endpoint is the same one used by the Chrome extension and WordPress plugin in production, so the contract is well-exercised rather than a marketing diagram.

When to pick which

Under 50 documents on a one-off basis with a human reviewer, the Bulk Scan UI is faster to set up and gives visual triage. Over 50 documents, recurring jobs, or anything you want logged to a database, the REST endpoint is the right answer. Both paths draw from the same Business allowance, so an account never gets penalised for switching between them mid-week.

Four shapes of bulk

Where the queue and the batch endpoint each earn their keep.

The four workflows we see most often. The UI handles the first two, the REST endpoint handles the second two, and the workspace audit log holds the trail for all four.

Agency content QA

An agency editor receives ten to forty drafts a week from freelancers and needs a defensible QA step before invoicing. Open Bulk Scan, drag the week's drafts in, sort by Authenticity Score, and paste per-piece result links into the project tracker. Workspace seats let two editors share one billing and one history, and a multi-client setup keeps each client's history in its own workspace so the audit trail stays clean per account.

Teacher class scanning

An English department processes 200 essays a term across four sections. The class submission folder gets dragged into Bulk Scan; ZIP upload handles a whole section at once. The queue sorts by score so the teacher reviews the highest-risk band first. White-label PDF export captures each result as an academic-integrity record, and the workspace audit log carries who scanned what and when as the artifact the dean asks for.

Content platform moderation

A user-generated content platform screens every submission before publish via the REST batch endpoint. The publish hook sends an array of pending posts, the response comes back with score, band, and highlights per post, and posts above the AI-likelihood threshold soft-block for human review. Webhook delivery handles the asynchronous case when the moderation queue spikes during product launches or seasonal traffic.

AI-content audits at scale

An SEO team or an editorial group audits a content archive for AI-written copy that pre-dates the team's quality bar. The audit script reads the sitemap, fetches each post, and batches them through the REST endpoint overnight, up to 500 files per call. Results land in a warehouse table with score, band, and per-piece highlights; dashboards aggregate by writer, client, and project. Flat allowance pricing fits the steady nightly volume; credit pricing would burn the budget on this shape.

11 file formats

Drop in whatever shape the documents arrive.

The Bulk Scan UI accepts paste, drag-and-drop, and ZIP archives. The unified extractor handles 11 file formats through officeparser, the same library the in-app /api/file/extract endpoint runs in production.

Documents and rich text

.docx for Microsoft Word, .pdf for native PDFs (image-only PDFs need OCR first), .rtf for rich text, .odt for OpenDocument, and .epub for ebooks. The extractor pulls clean body text from each one, drops document metadata, and queues the result as if it had been pasted in by hand.

Plain and web text

.txt for plain text, .md for Markdown, and .html for web pages. Markdown formatting is stripped before scoring so the classifier sees prose rather than syntax. HTML tags are stripped the same way before the text reaches the model.

Office spreadsheets and slides

.pptx for PowerPoint decks (slide text is concatenated in slide order), .xlsx for spreadsheets (cell text is concatenated across sheets), and .csv for tabular exports. Useful when student submissions or client deliverables arrive as decks rather than essays.

ZIP archives, up to 500 files

Drop a ZIP of any combination of those formats and the queue unpacks them document by document, with a Business cap of 500 files per batch. A class folder of 30 essays in mixed .docx and .pdf becomes 30 queue rows in one upload; a content archive of 500 articles becomes 500 rows ready for sorting.

Five-band triage at scale

The Authenticity Score, sorted for bulk reviewers.

Bulk results are useless without a triage rule. The Authenticity Score collapses to one of five bands and the queue sorts by band so you can act on the extremes first.

Original (80 to 100)

Clearly human writing. Pass through without further review. In a class of 30, expect 12 to 20 documents in this band on average. These rows can be dispositioned in bulk without opening the highlight view.

Mostly Human (60 to 80)

Human writing with some AI-assistance markers or polish. Skim, do not investigate. The triage rule is to scan the per-piece summary, not to open every row.

Mixed (40 to 60)

The genuine borderline. Open the sentence-level highlights to see which paragraphs lean AI, then make a judgment call. This is the band where the bulk workflow saves the most time, because the evidence comes pre-marked.

Likely AI (20 to 40)

Strong AI signal with some human intervention. Worth a closer look before any action, and worth a per-piece PDF export if the consequence is academic-integrity or contractual.

AI Generated (0 to 20)

High-confidence AI output. Cross-check with a second detector if the consequences are heavy, then act. In a slush pile of 50 submissions, expect 3 to 8 rows here; these are the ones the bulk workflow exists to surface.

Workspace + audit log

Five seats, shared history, multi-client setup.

Bulk reviewers are usually a department, an agency pod, or an editorial board. Multiple seats sharing one billing relationship with their own logins, shared history, and role-based access is table stakes by 2026. The audit log on top is the artifact procurement asks for.

Five workspace seats

Business includes five seats by default. The owner pays once, invites members by email, and every seat draws from the shared character allowance and sees the shared scan history. Roles cover owner, admin, and member with role-based access on billing and seat management. The team management module shipped in May 2026 with invite, join, and remove flows already in production.

Multi-client workspaces

Agencies running multiple clients can keep separate workspaces under one Business account so each client's history lives in its own audit trail without spilling across teams. The dashboard switches workspaces from a single dropdown and the REST endpoint scopes batch calls to the workspace key. One billing, many client trails.

Workspace audit log

The audit log records who scanned what and when, scoped to the workspace. Each row carries the scanner, the document title or hash, the score, the band, and the timestamp. This is the artifact institutional buyers and procurement teams ask for, and it is the difference between five personal accounts and a real team workflow that survives an HR audit or a client dispute.

White-label PDF reports

Per-scan PDF export with custom branding ships on Business, suitable for client deliverables and academic-integrity files. The PDF carries the score, the band, the sentence-level highlight overlay, and the per-model confidence numbers. A shareable result link lets a recipient open the same view as the reviewer without an account.

FAQ

Bulk reviewer questions, answered.

How many files can I scan per batch with the bulk AI detector?
Business unlocks 500 files per batch through the Bulk Scan UI on a 5-seat workspace with one shared queue and one audit log. The REST batch endpoint at api.textsight.ai/scan/v1/batch takes an array of documents in a single request and returns an array of scored results, which is faster than fanning out one call per document. Business includes 150,000 AI rewriter words per month alongside unlimited detection scans on a flat monthly allowance, so the practical ceiling is throughput rather than credits.
Which file formats does bulk upload accept?
The unified extractor handles 11 file formats through officeparser: .docx, .pdf, .txt, .md, .rtf, .odt, .epub, .html, .pptx, .xlsx, and .csv. Drag-and-drop, paste, and ZIP archives are all supported. The same /api/file/extract endpoint that powers the in-app Bulk Scan view runs here, so format coverage matches the live product line by line. Image-only PDFs need OCR before scoring; native PDFs work cleanly.
Is there a REST batch endpoint for the bulk AI detector?
Yes. POST an array of documents to api.textsight.ai/scan/v1/batch with an Authorization Bearer header and the same classifier that powers the dashboard returns an array of scored results in one call. Each response carries the Authenticity Score, the band label, per-sentence highlights with start and end offsets, and per-model confidence. Webhook callbacks are supported on Business so a CMS, LMS, or moderation pipeline can submit a batch and forget the connection.
Does Business include an audit log and multi-client workspaces?
Yes. Business ships a workspace audit log that records who scanned what and when, scoped to the team. Five seats per workspace share queue history and role-based access covers owner, admin, and member roles. Agencies running multiple clients can keep separate workspaces under one Business account so each client deliverable lives in its own audit trail without spilling history across teams. The audit log is the artifact procurement asks for.
How much does the bulk AI detector cost?
Business is $39.99 monthly or $29.99 monthly when billed yearly, which saves 25 percent. The tier includes the Bulk Scan UI with 500 files per batch, the REST batch endpoint, webhook callbacks, 5 workspace seats, the audit log, white-label PDF reports, and 150,000 AI rewriter words per month. There is no contact-sales gate. Pro at $19.99 monthly or $14.99 yearly covers unlimited single-document scans but does not include the batch endpoint or workspace seats.
Can teachers scan a whole class in one batch?
Yes. A class of 30 essays at 800 words each is roughly 144,000 characters and fits inside a single ZIP upload to the Bulk Scan UI on a Business workspace. Sort the queue by Authenticity Score to triage the highest-risk submissions first, then open any row for the sentence-level highlight view. White-label PDF export captures each result as an academic-integrity record, and the audit log holds the trail the dean asks for.
Will batch scans return a different score than single scans?
No. The same classifier runs whether a document arrives through the dashboard single-scan box, the Bulk Scan queue, the REST batch endpoint, the Chrome extension, or the WordPress plugin. A 500-document batch produces the same Authenticity Score, the same sentence-level highlights, and the same per-model confidence numbers you would get scanning each document by hand. There is no quality tier on the classifier itself.
What happens if one file in the batch errors out?
The batch keeps running. Errored rows flag in red with a short reason such as corrupt PDF, password protected, or text too short to score. The rest of the queue completes normally and you can retry only the failed rows from the same table without re-uploading the entire folder. The CSV export and REST response also include a status field so downstream tools can filter errored entries cleanly.
Related

More bulk and detection.

Clear the whole queue. Free to start.

Bulk Scan UI with ZIP upload across 11 formats, 500 files per batch. REST batch endpoint with webhook callbacks. Five workspace seats with shared history, multi-client support, and a workspace audit log. Flat allowance pricing that gets cheaper per-piece as volume rises.

Start bulk scanning free See Business pricing
500 files/batch · REST API · 11 formats · 5 seats · Audit log