rightaichoice/ai-tool-redirects-2026

AI Tool Redirects 2026

191 AI tools whose listed website now redirects to a different registered domain — and not one of them would be flagged dead by a standard liveness check.

Scanned 2026-08-21. 173 of the 191 return HTTP 200; the remainder return 403, 301/308 or 503 — every one of which a conservative liveness checker treats as alive.

A companion to The Product Hunt Graveyard, which measures tools that died. This one measures something harder to see: tools that changed identity while continuing to look perfectly healthy to any automated check.

Published by RightAIChoice.

DOI (archived version of record): 10.5281/zenodo.22036427 Companion DOI (mortality data): 10.5281/zenodo.21598672

---

Why this dataset exists

Almost every liveness checker — ours included, until this scan — treats anything that is not a hard failure as alive. HTTP 200 obviously, but also 403 (bot wall), 503 (temporary) and 3xx (redirect), because treating those as deaths produces constant false positives.

A cross-domain redirect returns 200. So when a product is acquired and retired, or its domain is sold, or the company rebrands, the check passes. The tool scores healthy, gets a recent verification date attached, and that date makes the wrong answer more convincing rather than less.

We found this in our own catalogue and published the data rather than quietly fixing it.

Two examples from the scan:

• reach-best — reachbest.co now redirects to an online-gambling site. Our own system rated it viability 73, "verified".
• monkeylearn — redirects to medallia.com after acquisition; the standalone product is discontinued. Rated viability 87.

Both were unpublished from our catalogue on 2026-08-20, the day before this scan ran. They appear in this dataset because the scan captured the state that let them through.

What is in it

191 rows, one per tool, from a full probe of all 8,164 published tools in the RightAIChoice catalogue on 2026-08-21. Every figure in this card describes that date and is not updated afterwards.

column	description
slug	Tool identifier
name	Tool name
listed_url	The URL on record before the scan
resolved_url	Where it actually landed after following redirects
listed_domain	Registered domain of listed_url (eTLD+1)
resolved_domain	Registered domain of resolved_url
redirect_class	acquirer · unrelated_domain · similar_domain
http_status	Final HTTP status after following redirects. 200 (173), 403 (9), 308 (7), 503 (1), 301 (1). None of these is treated as dead by a conservative checker — that is the point
viability_score_at_scan	Our health score at scan time, 0–100
last_verified_at	When our pipeline last checked it
scanned_at	2026-08-21

Breakdown

class	count	meaning
acquirer	14	Redirects into a large acquiring company. Highest likelihood the standalone product was retired
unrelated_domain	88	Redirects to a name with no textual relationship. Mix of clean rebrands and acquisitions
similar_domain	89	Domain moves such as .co→.com or .app→.ai. Usually benign, but the stored URL is stale

Viability scores at scan time: minimum 25, median 69, maximum 95. Every one of these was passing.

Honest limits — read before citing

A redirect is not a death certificate. It proves identity moved, not that the product ended. Many of these are healthy rebrands — codeium.com → windsurf.com is a live, thriving product.

Distinguishing "rebranded", "acquired but alive" and "acquired and killed" requires a human looking at what is actually on the other end. This dataset deliberately does not make that call, because the redirect alone cannot support it.

redirect_class is heuristic. acquirer is matched against a hand-written list of large companies, so it under-counts. similar_domain uses string similarity below 0.45 and will occasionally misfile an unrelated rebrand.

One methodology bug, disclosed. The first pass used lstrip("www.") to strip subdomains — which removes characters, not the prefix, so windsurf.com became indsurf.com. It produced one false positive. The matcher was corrected and the published figures come from the corrected run. Recorded here because a dataset about undetected failure should not hide its own.

Also observed but not included: 262 of the 8,164 tools were unreachable at scan time (mostly DNS or timeout). Unreachable at one moment is not evidence of anything, so those rows are excluded entirely rather than counted as dead.

Method

One GET per tool with redirects followed, browser user-agent, 20-second timeout, 24 concurrent workers. The registered domain (eTLD+1, with a multi-part suffix list for .co.uk-style domains) of the final URL is compared against the listed URL's. Any difference is a row.

License & citation

CC BY 4.0 — free to use, including commercially, with attribution.

Verma, Tanmay. (2026). AI Tool Redirects 2026: Products That Changed
Identity While Still Passing Liveness Checks (RightAIChoice) [Data set].
Zenodo. https://doi.org/10.5281/zenodo.22036427

Related

• The Product Hunt Graveyard 2026 — 2,291 top launches probed; 24.4% of the 2,066 with a definitive verdict are dead
• Zenodo 10.5281/zenodo.22036427 — the archived, citable version of this dataset
• Zenodo 10.5281/zenodo.21598672 — the companion mortality dataset
• Methodology