A native LTI 1.3 integration for Canvas SpeedGrader is in planning, targeting Q4 2026. AI-likelihood scoring and sentence-level highlights would load alongside the rubric pane, private to the instructor by default, FERPA-scoped, and configurable for self-hosted Canvas. The shipping date is contingent on waitlist demand. If your institution wants a seat at the design table, join below.
We picked Canvas first because Instructure published clear LTI 1.3 docs, the SpeedGrader pane has the right hooks, and the institutional pain we hear most often is shaped like one of these five.
Today an instructor opens SpeedGrader, copies the submission, opens a second tab, pastes into TextSight, reads the score, switches back, and types a manual comment. That round trip costs roughly forty-five seconds per submission. Across a 120-student composition class with weekly drafts, that is ninety minutes of pure tab-switching every grading cycle. A pane inside SpeedGrader collapses it to one click.
Detection is most useful when the instructor can see which specific lines flagged while reading the rubric criteria. A native pane lets sentence-level highlights load over the same submission text Canvas already renders, so the instructor reads once instead of twice. The current paste workflow loses that alignment because the highlighted sentences live in a different DOM.
The instructor sees the score in the SpeedGrader pane; the student does not see it on the submission detail page. That mirrors how rubric-internal notes already work in Canvas. The current paste workflow has no way to enforce that boundary, which is why some institutions disallow tab-side detectors during academic-integrity review.
A native integration that scans in transit and does not retain student text addresses the single biggest objection institutional buyers raise: "we cannot send raw student work to a vendor we have not signed a DPA with." The roadmap below commits to a no-retention scan path, hashed audit trail, and signed Data Processing Addendum before any institutional rollout.
Roughly twenty per cent of Canvas-using institutions we hear from run self-hosted, and Chrome-extension-only workarounds leave them out. Building on LTI 1.3 (rather than a Cloud-only API) means the same install works on Instructure-hosted Cloud, on self-hosted instances behind a university VPN, and on Canvas Free for Teachers used by individual classrooms.
Concrete buckets, not aspirational vapour. Each card is a thing we have either started or scoped well enough to estimate. Target ship for the full Canvas integration is Q4 2026, conditional on the waitlist hitting roughly fifty institutional signups by August 2026.
A documented prototype that registers TextSight as an LTI 1.3 tool inside our Canvas test sandbox, completes OIDC sign-in, and reads a sample submission via Names and Roles. No production traffic.
Collect institution names, instructor headcount, and Canvas environment shape. Top fifty by headcount get invited to a design-partner programme and an early closed beta.
An embeddable pane that loads against a single submission, calls our existing scan API, and returns score plus sentence highlights. Internal-only build, against the test sandbox, no student-facing surface yet.
Publish a draft Data Processing Addendum, subprocessor list, and retention policy specific to Canvas student submissions. Open for institutional review before code ships.
Production install on partner Canvas tenants. Capped at five institutions. Real student submissions, instructor-private results, no public marketing of the partner names without consent.
Public Instructure listing, self-serve install for any Canvas tenant, and pricing finalised on the TextSight Business plan with an institutional Enterprise tier above fifty instructor seats.
Dates are best-effort, not contractual. If the waitlist signal is weak by August 2026, we will reshape the roadmap rather than ship a half-baked Q4 release.
FERPA compliance is a shipping requirement, not a marketing line. The design below is the planned architecture, open for institutional review during the Q3 2026 closed-beta window.
When an instructor clicks Scan inside the SpeedGrader pane, the submission text is sent to TextSight over TLS, the AI-likelihood model runs against it, and the score and sentence-level highlights are returned. The raw submission text is not retained on TextSight servers beyond the scan request. The only persisted artefact is the score and the per-sentence highlight indexes, scoped to the instructor's audit window.
Institutions can set the audit window from a 90-day default down to seven days for tighter retention policies. After the window expires, the score and highlight records are purged. The hashed submission identifier remains in the audit log so the institution can prove a scan happened without our side holding the underlying text.
A signed Data Processing Addendum will be published before the general-availability ship. The subprocessor list (cloud hosting, model inference, error logging) will be enumerated with locations and the data each one processes. Institutional buyers can request a draft DPA today by emailing support@textsight.ai with a one-line statement of who at the institution will sign.
Scores are not surfaced to students by default. An instructor-level toggle exposes the score and highlights inside the student-facing submission view, the same way rubric feedback already surfaces. Institutional admins can lock the toggle so individual instructors cannot override the policy. No score is ever shared with the student before submission unless the instructor has explicitly enabled a pre-submission self-check workflow.
This is not a covert detection layer. Students will be able to see, via the standard Canvas LTI consent flow, that a TextSight scan was performed on their submission. That is a feature, not a leak: covert detection erodes student trust and creates appeal-process problems for the institution. Transparency on which submissions were scanned, with what scope, is the design default.
Saying no publicly is how we stay honest about the roadmap. These items are not in scope for the Q4 2026 ship, and several are not planned at all.
TextSight's Plagiarism Risk score is a web-search-driven similarity signal, not a closed academic submission database like Turnitin's. The Canvas integration will surface our existing risk score, but it will not crawl Canvas itself to build a cross-class submission index. If your institution needs a closed-database plagiarism workflow, Turnitin remains the right tool. We are not pretending to replace that.
The pane returns a score and highlights. It does not make grading decisions, it does not gate submission acceptance, and it does not file academic-integrity reports automatically. That is on purpose. AI-detection scores are probabilistic, false positives exist, and the instructor stays in the loop on every consequential decision.
TextSight's AI rewriter exists in the web app and the browser extension. It will not ship inside the Canvas pane for student-facing surfaces, because giving students an in-LMS rewriter to defeat detection is a conflict of interest the integration is not going to take on. Instructors get a one-click rewriter for their own demo cases on the web app side.
Each LMS gets its own integration page, its own LTI configuration, and its own honest roadmap card. Canvas is first because the LTI 1.3 docs are clearest and the design-partner conversations have been densest. Blackboard, Moodle, Schoology, and Brightspace stub pages are tracked separately, none of them are shipped today.
The waitlist is the demand signal. Roughly fifty institutional signups by August 2026 unlocks the Q4 ship. Joining is non-binding on both sides, and top-fifty signups get a design-partner invitation.
Submitting opens your mail client addressed to support@textsight.ai. Please include Canvas environment (Cloud or self-hosted) and rough instructor headcount in the body. No tracking pixel on this form. No automated signup, on purpose.
If your question is not below, email support@textsight.ai. Replies go to a human inside our team, not a routing bot.
Not yet. The Canvas integration is in planning, with a target ship of Q4 2026. The integration is gated on waitlist demand: if enough Canvas-using institutions sign up by August 2026, we commit to ship that quarter. Until then, instructors can paste assignment submissions into the TextSight web app or browser extension to scan them, and mirror the result back into SpeedGrader as a manual comment.
Instructor opens a Canvas SpeedGrader submission; a TextSight pane appears alongside the rubric; the submission is scanned in under three seconds; the AI-likelihood score and sentence-level highlights load in-place. The score is private to the instructor by default. Students do not see it unless the instructor explicitly shares the report through the SpeedGrader comment thread.
LTI 1.3 with Deep Linking. That is what Canvas Cloud, Canvas Free for Teachers, and self-hosted Canvas instances all support without admin-side custom installs. LTI 1.3 also gives us OIDC-based single sign-on and Names and Roles Provisioning, which we need to honour instructor-vs-student visibility. A Chrome-extension-only path leaves self-hosted users out, so we are not building that as the primary surface.
FERPA compliance is a shipping requirement. The planned design: student submissions are scanned in transit, the text is hashed for audit, and the raw submission is not retained on TextSight servers beyond the scan request. Score and highlight records persist for the instructor's audit window, default 90 days, configurable down to seven. We will publish a signed Data Processing Addendum and a subprocessor list before launch.
Pricing is not finalised. The likely shape: included on TextSight Business at $39.99 monthly ($29.99 on annual billing) per instructor seat for small deployments, with an institutional Enterprise tier above fifty instructor seats negotiated per-deal. Canvas itself bills you separately through Instructure. There will be no add-on cost to Canvas; the integration runs against your TextSight account.
That is an instructor-controlled toggle, default off. Surfacing the score before submission turns the detector into a coaching tool students can game. Instructors who want a pre-submission self-check workflow can flip the toggle on, and students will see their score the same way they see SpeedGrader rubric feedback. Both paths will be supported. Neither is the default for the institution.
Email support@textsight.ai with your institution name, Canvas environment (Cloud or self-hosted), instructor headcount, and rough timeline. The waitlist is non-binding on either side. Joining helps us prioritise: if your institution is in the top fifty by headcount, we will invite you to the design-partner programme and ship a closed beta to your account before the general release.
Sibling pages across our LMS stubs and educator workflows. None of the LMS-side integrations are shipped yet; the educator-facing tools are.
The umbrella page for everything TextSight does on the educator side today: AI detection, sentence-level highlights, classroom self-scan workflows, and the .edu Pro pricing.
Classroom-scale framing of the same tooling: how to run TextSight against essays, drafts, and short-response submissions during a grading cycle.
Sister stub page for Blackboard. Same shape as this Canvas page, separate waitlist signal. Q1 2027 target if Canvas ships on time.
Stub page for Moodle, focused on the self-hosted European university context. Plug-in repository path is being scoped separately.
K-12 LMS stub for the Google Workspace ecosystem. Lightweight grading plus Drive-native submission workflows; tracked separately from Canvas.
The benchmark protocol, test-set composition, and false-positive numbers behind the detector that would power the planned Canvas pane.
Join the waitlist if you want the native flow when it lands. Until then, your instructors can use the free TextSight web app or browser extension to cover the same workflow manually.