Cases › produced harmful content › LR-2026-0887
Podqi investigation: ChatGPT, Gemini and Perplexity shopping answers included risky sites
- Who was hit
- 718 risky domains (351 fake sites and 367 inactive sites) surfaced across the 6,795 responses analysed
- How it was found
- Third party
- Reported
- 13 Nov 2026
- Severity
- Minor
- ChatGPT, Gemini and Perplexity answered shopping questions about 151 brands. Podqi tested them in September.
- The services included risky shopping sites in answers: fake malls, counterfeit sellers, re-registered domains and dead sites. Most such responses recommended the site as a place to buy.
- The investigation found 718 risky domains, 351 fake and 367 inactive, in about 9 percent of the 6,795 responses.
- Podqi, an intellectual property protection firm, found the risky sites in its investigation of the three services. The Verge reported it.
Source: 디지털투데이
What was done
OpenAI said it will improve safety measures for shopping search results, and Perplexity said it continues to focus on accuracy.
Other stated action (Developer or vendor) · Other stated action (Developer or vendor)
Fix reported: “OpenAI said it will improve safety measures for shopping search results.”
Risks Harmful content and jailbreaks (75 incidents)
No source states: AI company and model, money lost or records, tools it had, company using the AI, sector.
Evidence
Sources (1)
- AI chatbot shopping recommendations include risky sites, hundreds identified · 디지털투데이 · 2026-11-13
Show every field, with its source and grade
What happened
Earliest source we hold
2026-11-13
D inferred by us from the described behaviour, reason shown
Who caught it
an outside party
D inferred by us from the described behaviour, reason shown
What was done
OpenAI said it will improve safety measures for shopping search results.
D inferred by us from the described behaviour, reason shown
What went wrong
produced harmful content
D inferred by us from the described behaviour, reason shown
Impact
Severity
minor
D inferred by us from the described behaviour, reason shown
Who was affected
718 risky domains (351 fake sites and 367 inactive sites) surfaced across the 6,795 responses analysed
D inferred by us from the described behaviour, reason shown
Why this severity
Risky sites appeared in about 9 percent of responses in one outside evaluation; no loss or third-party harm is stated.
D inferred by us from the described behaviour, reason shown
Context
Base count
6,795 units, 2 rounds in September
D inferred by us from the described behaviour, reason shown
LR-2026-0887 · version 2 · published 2026-11-13 after automatic checks; not yet checked field by field by a person