Be My AI reads airport gate numbers and a check for blind users

AI system

System: Be My AI (large-multimodal-model image description in the Be My Eyes app)
Model: GPT-4.
Release: Exact release not established.
Application developer: Be My Eyes (feature built on OpenAI's GPT-4)
Role: The model described images for people with impaired vision.
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Smartphone photograph submitted to Be My AI by a blind or low-vision user. Task in a familiar or low-stakes setting. User interprets a text description and acts on it.
Human oversight: Not assessed
Relation to AI: The multimodal model generated the descriptions participants acted on. No independent per-task accuracy check was made. Participants themselves reported risk (P9: “I knew there was a risk”).
AI contribution: The multimodal model generated the descriptions participants acted on. No independent per-task accuracy check was made. Participants themselves reported risk (P9: “I knew there was a risk”).

Benefits

Outcome identifiers: Not assessed
Benefit: Peer-reviewed interview studies record blind users completing everyday tasks with Be My AI, e.g., airport gate numbers, a handwritten check.
Category: Improved Accessibility
Status: Not assessed
Beneficiary groups: People.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: Not assessed
Comparison: Assumed alternative: asking a sighted family member or friend to check a photo
Results: Assumed comparison. Reference: Before GenAI, I would often send it to my mom or friend to check for me. Result: Now I Ask ChatGPT.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: Be My Eyes (feature built on OpenAI's GPT-4). Operator: Be My Eyes (consumer app used by study participants).
Jurisdictions: Developer: United States. Operator: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: The model does not hallucinate in the specific scene (both studies document hallucinations). User cross-checks high-stakes outputs such as medication or finances.
Replication attempts: 1 follow-up entries. Benefit recurred: 1.
Later findings: 30 June 2025. An independent interview study run from November 2024 to January 2025 recorded blind participants still using Be My AI for everyday reading tasks such as a handwritten check.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0034-source-0001–0003
Citations: Numbered references below.
Language: Not assessed
Source relationships: Independent
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People.
Depth: People: 1.
Reasons: People: Everyday tasks (gate numbers, orientation, locating a dropped item, reading a check) completed independently by identifiable participants. Small, self-reported, and the same papers document frequent errors.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently studied

Entry details

Kind

Not assessed

Related entries

Same stakeholders and task: AI describing visual content for blind users; Be My AI read gate numbers and a check, while AI accessibility overlays mislabelled web content, misleading blind and visually impaired people such as Jakob Rosin.

Field

Accessibility / visual assistance. Blind and low-vision users.

Purpose

Be My AI was built to describe photographed scenes and objects to blind and low-vision users. These task completions were that purpose. Xie et al. state that Be My AI is built on OpenAI's GPT-4 models.

Dates

Date basis unassessed: 2023 (precision unassessed). Updated: 21 October 2026.

Stage

Not assessed

Version

SHA-256: 6ddf264b722b…

References

Other identifiers: AI Incident Database: 830.
Category mappings: Not assessed
Brief: Not assessed
Revisions: 6 revisions. Latest: 21 October 2026.

Sources

Peer-reviewed study Independent author

Xie J, Yu R, Zhang H, Billah SM, Lee S, Carroll JM (2025). Beyond Visual Perception: Insights from Smartphone Interaction of Visually Impaired Users with Large Multimodal Models. Proc. CHI 2025 (PMC12338113).
Preprint Independent author

Sharma T, Tseng Y-Y, Zhang L, Ide A, Mack KA, Findlater L, Gurari D, Wang Y (2025). 'Before, I Asked My Mom, Now I Ask ChatGPT': Visual Privacy Management with Generative AI for Blind and Low-Vision People. arXiv:2507.00286 (published ASSETS 2025).
Other Independent author

Penn State Pure research record (2025). Publication venue entry for Sharma et al.