A deaf student sitting in a mainstream classroom with auto-captions running on a screen is technically receiving captioning. Whether that constitutes equal access depends on what those captions are producing.

The difference between real-time CART captioning and AI-generated auto-captions comes down to accuracy: human CART providers consistently achieve 98% and above, while AI auto-captions drop to 70–80% under favorable conditions and lower when the classroom environment gets harder. That gap is the center of most captioning conversations between families and schools.

The Accuracy Gap and Why It Matters in Practice

CART, or Communication Access Realtime Translation, is a service provided by a trained human stenographer who transcribes spoken content verbatim in near-real time. Professional CART providers run at speeds exceeding 225 words per minute with accuracy consistently at or above 98%.

AI auto-captions, whether through Google's Live Transcribe, Microsoft Translator, or platform-specific features, operate through automatic speech recognition. Under good conditions, with a single speaker, minimal background noise, and standard speech patterns, accuracy can reach 80–85%. In a typical classroom with multiple speakers, specialized vocabulary, accented speech, and ambient noise, accuracy drops further.

The gap matters most in the moments that count: a lecture on a technical topic where a misheard term produces a wrong answer on an exam, a class discussion where one missed word changes the meaning of the next three sentences, a lab instruction where accuracy affects safety. The 2014 Joint Committee on Infant Hearing's position on access for students with hearing loss specifies full access to classroom communication, not approximate access.

When Auto-Captions Are Adequate

Auto-captions are not uniformly inadequate. There are classroom situations where they provide meaningful access:

• Recorded lecture videos, where the student can pause and rewind, and where caption errors are reviewable in context
• Teacher monologue in a low-noise environment with standard vocabulary and a single speaker
• Informal review or study sessions where context compensates for occasional errors

Where they consistently fall short is live instructional settings with multiple speakers, dynamic discussion, or specialized terminology. The school default of using auto-captions for all captioned instruction reflects a cost calculation, not an assessment of whether the access provided is legally or educationally sufficient.

Requesting CART in an IEP or 504 Plan

CART is an assistive technology accommodation. For students with IEPs, the request belongs in the AT section. For students on 504 plans, the 504 coordinator manages the response, and the plan must name the accommodation in writing.

The accommodation request should include:

• Real-time CART captioning for all live classroom instruction
• Remote CART as an acceptable delivery method for hybrid or virtual settings
• The party responsible for locating and contracting with a CART provider
• A backup protocol for when a CART provider is unavailable

Most families who request CART for the first time are walking into a conversation where the school has already settled on auto-captions as sufficient. Schools often resist CART requests on cost grounds. The ADA's standard is not whether the accommodation is expensive but whether it is necessary for equal access. The basis for that argument is documentation: if auto-captions are failing in live instruction settings, request a communication access audit from the school's educational audiologist. If the district does not have one, the state Department of Education's deaf and hard-of-hearing specialist can often conduct or commission an assessment.

Remote CART and Hybrid Options

Remote CART connects a stenographer working from another location to the student's screen via a secure web link. Transcript quality matches in-person service, and remote delivery has significantly expanded provider availability for schools in rural or underserved areas.

Verbit, Caption Access, and Ai-Media all offer remote CART for educational settings. The National Court Reporters Association maintains a directory of CART-certified providers at ncra.org.

For situations where full-price CART is not feasible, hybrid models that combine automatic speech recognition with human editing, such as Ava, can approach CART-level accuracy at lower cost. They introduce a slight delay and work best in situations where near-CART accuracy is needed but full service is not achievable within the current budget.

A student who cannot follow classroom instruction because captions are missing critical content is losing educational material every day. The school's obligation to provide effective captioning does not change based on which option is cheapest to deploy. A classroom observation, a documented accuracy gap, and a written accommodation request are the materials that argument needs, and all of them are within reach.