GPT Researcher
- Category
- Research
- Type
- open-source
- Source
- Project repository
- Last reviewed
- 2026-10-16 (what this means)
Summary
GPT Researcher is an open-source research agent that writes cited reports. In independent evaluations (DeepResearchGym, ICTIR 2026; Deep Research Comparator, WWW 2026 Companion; TTD-DR preprint), results ranged from comparatively strong on citation quality to lower than a commercial deep-research system in head-to-head comparisons. Results vary with the AI model and search provider used. Informational only; not a guarantee, endorsement, or certification.
Overall score: 53 · Evidence grade: B- · Independent sources: 3 performance, 7 other (vulnerability records)
Factor breakdown
| Factor | Weight | Score | Note |
|---|---|---|---|
| Completion rate | 30% | No published completion-rate figure. Independent studies show complete cited reports, with mixed head-to-head results. Judgment call. | |
| Disputes | 20% | No user disputes found. Public databases list 7 security advisories; one is fixed per the maintainer. Judgment call. | |
| Adherence to limits | 25% | Configurable limits are documented by the project. The self-hosted server ships without built-in authentication per its security policy. Judgment call. | |
| Evidence quality | 25% | Three independent performance evaluations (one main-conference paper, one companion-track paper, one preprint) plus public vulnerability databases. Judgment call. |
Overall = 0.30×55 + 0.20×40 + 0.25×50 + 0.25×65 = 53.25, shown as 53. See the methodology.
Security
Public vulnerability databases (NVD and GitHub Advisory Database) list 7 security advisories (2 Critical, 3 High, 2 Medium). The maintainer has fixed one; several others concern the self-hosted server, which the project's security policy says ships without built-in authentication and is intended for trusted environments only. Status may change; see linked sources.
Uncertainties
- Results depend on the AI model and search provider used; the evaluations used differing, mostly 2025-era configurations.
- Fix status of some advisories was not independently verified.
- One independent evaluation (TTD-DR) is a preprint by authors comparing against their own system.
- No independent completion-rate or cost data was found.
Sources
- DeepResearchGym benchmark results (arXiv v3) — independent performance, accessed 2026-10-16
- DeepResearchGym ICTIR 2026 camera-ready — independent performance, accessed 2026-10-16
- DeepResearchGym ACM DL record — independent performance, accessed 2026-10-16
- Deep Research Comparator human-preference evaluation — independent performance, accessed 2026-10-16
- Deep Research Comparator ACM DL record (WWW 2026 Companion) — independent performance, accessed 2026-10-16
- TTD-DR preprint; GPT-Researcher as a baseline — independent performance, accessed 2026-10-16
- NVD: XSS (Medium) — independent other, accessed 2026-10-16
- NVD: XSS (Medium) — independent other, accessed 2026-10-16
- NVD: code injection (High) — independent other, accessed 2026-10-16
- NVD: missing authentication (High) — independent other, accessed 2026-10-16
- NVD: SSRF (High) — independent other, accessed 2026-10-16
- NVD: WebSocket code execution (Critical) — independent other, accessed 2026-10-16
- GitHub Advisory: CVE-2025-65720 (Critical) — independent other, accessed 2026-10-16
- Maintainer sanitization fix (PR #1820) — vendor or maintainer, accessed 2026-10-16
- Project security policy / threat model — vendor or maintainer, accessed 2026-10-16
- Maintainer self-reported benchmark — vendor or maintainer, accessed 2026-10-16
- Documented configuration limits — vendor or maintainer, accessed 2026-10-16
- Project repository — vendor or maintainer, accessed 2026-10-16