Why is my radiology software missing from AI answers?
The question about what to look for in a new radiology PACS named none of the 24 checked vendors across fifteen valid answers captured on 22 September 2026. Across the full 150-answer radiology category, only Canon Medical Informatics and Hyland were never named. That difference makes the question essential to any absence report. The collection does not identify why a supplier was omitted.
What we measured
Measured. The September radiology and medical-imaging panel contains ten fixed questions with three scheduled answers on ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode. All 150 observations were valid. The table shows the complete category-wide zero set, rather than the larger set left unnamed on the PACS-criteria question.
Both zero-of-150 rows have a descriptive Wilson 95 percent interval of 0.0 to 2.5 percent. Captures are dated 22 September 2026; publication was on 23 September. Name matching records appearances under a stated rule, without testing an imaging system or the quality of clinical work.
| Checked vendor with no recorded names | Named in valid answers |
|---|---|
| Canon Medical Informatics | 0 of 150 |
| Hyland | 0 of 150 |
The criteria question is a fifteen-answer subset within the 150-answer category. Its absence of checked names does not establish that the answers failed to explain selection criteria. Assessing that would require reading the explanation, not just counting names.
Source lists are a separate observation. reddit.com appeared in 60 of 150 category answer lists and intelerad.com in forty. Those pooled host counts can help locate material for inspection, but neither one establishes the cause of a missing name or verifies a PACS function.
Every figure comes from the published method, v1.1.
What this means for you
Reasoned. Preserve the setting behind your concern: an outpatient centre, a group of imaging locations or a hospital team. A question asking for selection criteria should not inherit the same expectation of product names as an open product search. Find the actual retained question before classifying the observation.
Check the vendor identity and aliases used in the record. A checked zero and an unchecked company are different states. If the product was outside the twenty-four-vendor roster, this collection supplies no measurement for it. Do not turn that missing coverage into a zero.
For an error investigation, retain one claim about the named system and the dated source needed to check it. Resolve the exact product or edition before proposing a factual correction. These historical omissions do not demonstrate a product defect, establish system suitability or predict a naming change after a page edit.
Also asked as
Why is my radiology software missing from AI answers?
Related
- Radiology question and source records
- What a measured zero records
- What an own-domain citation records
- ChatGPT radiology observations
Next step
Read the imaging context and original question in Free 10-question check (one engine). Separate a missing vendor name from an unsupported product statement. No email is needed; runs are limited per network and per day, and by the site's daily allowance.
When primary product evidence resolves one factual error, Fix one answer, $490 can address that supported page correction. Keep the delivered fact and any later appearance as distinct records.
Claiming or correcting a vendor record is free and never depends on a purchase.
Prepared by the Broadcastwell team.
Page owner: Ananya Rao, AI Engineer. Corrections: hello@broadcastwell.com