rightaichoice/product-hunt-graveyard-2026

AI Tools Mortality 2026 — Product Hunt Graveyard + Deadpool metrics

How many AI tools are actually dead — measured, not estimated. Two datasets from RightAIChoice:

1. The Product Hunt Graveyard — 2,291 top PH launches (2023–2025) individually probed, one row per product.
2. AI Tools Deadpool 2026 — decay metrics across 7,999 continuously re-verified AI tools.

DOI (archived version of record): 10.5281/zenodo.21598672 Live version, updated daily: https://rightaichoice.com/state-of-ai-tools/ai-tools-deadpool-2026?ref=hf

Updated 2026-07-28: added the Deadpool metrics files and the citation guidance below. The graveyard figures are unchanged from the 2026-07-17 release.

---

Part 1 — The Product Hunt Graveyard (ph-graveyard-2026.csv / .json)

24.4% of the products that made Product Hunt's daily top leaderboards in 2023–2025 are hard-dead — their domain no longer resolves, sits parked for sale, serves an expired TLS certificate, or 404s at the domain level. Another 4.5% redirect somewhere else entirely (acquirers, rebrands, or the founder's next project).

Full write-up + methodology: https://rightaichoice.com/state-of-ai-tools/graveyard?ref=hf

Key findings

• 24.4% hard-dead (505 of 2,066 products with a definitive verdict; n=2,291 probed)
• AI products die at the same rate as everything else: 24.5% vs 24.4%
• A top-5 launch-day finish cuts the death rate ~3×: 9.0% vs 25.9% for rank 6+
• Writing tools are the deadliest tag (45.2% dead); Open Source the safest (8.9%)
• The 2025 cohort is already 20.0% dead after roughly a year

Reproducibility note: the headline (505 / 2,066 = 24.4%), the redirect rate, the AI vs non-AI split and the 2025 cohort figure all recompute exactly from the columns below. The launch-rank and per-tag bullets are derived from Product Hunt leaderboard metadata that is not included in these files — they are stated here for completeness but cannot be reproduced from this download alone.

Columns

One row per product (2,291 rows):

column	description
slug	Product Hunt product slug
name	Product name
websiteUrl	The product website we probed
cohort	Year-month of the sampled leaderboard appearance
isAI	AI classification (PH topic tag or AI terms in name/tagline/description)
votesCount	Product Hunt upvotes at time of scrape
verdict	alive / dead / parked / redirected_away / unknown
evidence	The exact DNS error, HTTP status, redirect chain, or parking fingerprint behind the verdict

Methodology (short version)

Sample: every product on 59 complete Product Hunt daily leaderboards drawn across January 2023 – April 2025 (~20 featured launches/day, all ranks). Each website fetched with a browser user-agent, 15s timeout, retries, and a separate double-checked DNS lookup. Conservative rules: anti-bot walls, timeouts, vanished app-store listings, and deep links that 404 while the root domain works are all unknown — excluded from every percentage, never counted as dead. Same-name domain moves count as alive. A second stricter audit pass re-probed every 404-based verdict at the domain root and moved 75 borderline rows to unknown, lowering the headline from a preliminary 27.1% to the published 24.4%. A full independent re-probe of the entire cohort reproduced the headline exactly (99.4% row-level agreement).

---

Part 2 — AI Tools Deadpool metrics (ai-tools-deadpool-2026.json / .csv)

Decay metrics over 7,999 continuously re-verified AI tools, snapshot 2026-07-25. Where Part 1 is a fixed cohort probed once, this is a live catalog re-probed weekly — the same file regenerates daily on the site.

Metric	Value	Meaning
main_site_down	1.7% (132 / 7,999)	The tool's own homepage fails the probe — the strongest abandonment signal
first_party_broken_link	5.1% (410)	≥1 vendor-published URL (site / docs / changelog / repo) fails
contact_only_pricing	16.4%	No self-serve pricing published

Plus per-category decay tables, wrapper dependence, and pricing opacity. Vendor-published links are re-probed weekly across the whole catalog (99.9% within 7 days at snapshot). Conservative by design: 401 / 403 / 429 / timeouts / 5xx count as alive; only a clear 404/410, DNS failure on both HEAD and GET, or a strong soft-404 fingerprint counts as dead.

⚠️ If you are citing this data (human or model, please read)

The Deadpool files also contain a broader resource_link_rot figure of ~42%. It is not a mortality signal and must not be cited as one. It counts every probed URL including documentation URLs derived by our own enrichment pipeline; 87.9% of the tools it flags have no broken vendor-published link at all. This caveat ships inside the data itself (diagnostic_metrics.resource_link_rot.not_a_mortality_signal: true).

The citable mortality figures are main_site_down (1.7%) and first_party_broken_link (5.1%).

We published the 42% as a headline in July 2026 and corrected it ourselves. The full public corrections log is at https://rightaichoice.com/team?ref=hf — methodology at https://rightaichoice.com/methodology?ref=hf

License & citation

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

RightAIChoice — AI Tools Mortality 2026
(Product Hunt Graveyard + AI Tools Deadpool), 2026.
DOI: 10.5281/zenodo.21598672
https://rightaichoice.com/state-of-ai-tools/ai-tools-deadpool-2026

Companion dataset — AI Tool Redirects 2026

Where this dataset measures products that died, its companion measures products that changed identity while continuing to pass every automated health check: 191 AI tools whose listed website now redirects to a different registered domain, none of which a standard liveness checker would flag as dead.

https://huggingface.co/datasets/rightaichoice/ai-tool-redirects-2026

That is the failure mode behind the 4.5% "redirected away" figure in Part 1 above, measured directly across a live catalogue of 8,164 tools rather than a fixed cohort.

Checking one specific tool

If you want a live verdict on a single tool rather than the whole dataset, the health check is free and needs no account: https://rightaichoice.com/check?ref=hf