Agent-Native Discovery: agent.reviews, "SEO for Agents," and the Fight Over What Tools Agents Recommend

Agent-Native Discovery: agent.reviews, "SEO for Agents," and the Fight Over What Tools Agents Recommend

A new layer is forming between tool vendors and agent-mediated decisions: agent-native reputation/discovery infrastructure. On October 7, 2026, "Agent.reviews – Where AI agents read and write reviews on tools" hit Show HN (72 pts, 49 comments). The reviews on it are written by users' agents, not humans; the founder (HN user screm) is associated with Armature (armature.tech, per the thread a YC-backed company whose pitch is: "Be the tool Claude Code chooses… With Armature get recommended and implemented for any user in any context. Then see what users do, and run evals so it keeps working." — quoted by commenter hypfer, who adds: "My agent called it SEO for agents.").

The strategic framing that makes this a market-map layer, from the founder:

"Claude Code for example picks Anthropic's code review tool in 80% of the cases. At some point this will probably face antitrust considerations but for the time 3rd-party vendors need to survive. We don't want a world where only labs survive, do we?" — screm

The HN thread functioned as a peer review of the model, and the failure modes it surfaced are the layer's structural problems:

• Gaming/injection surface: "I noticed lots of talk about privacy but this seems to be a prompt injection factory no?" — conception (compare Security & Prompt Injection Vulnerabilities in Agentic Research Loops)
• Grade inflation / broken proxies: "cryptography got 4.6/5 stars. That sounds pretty good, but then again YAML also got 4.6/5 stars. I'm thinking the agents are grading on a curve"¹ — themgt; the founder concedes "agents never put 1-star rating so we may need to normalize the scale" — screm. This is Goodhart's law arriving in tool discovery (AI Is Breaking the Proxies for Expertise: Goodhart's Law Comes for Research Output).
• Cold-start incentive problem: "What is the incentive for me to spend my tokens on submitting reviews?" — klntsky
• The monetization endgame everyone expects: "It's very easy to see this is going to be pay-to-rank-higher as the next move once some critical mass of people starts pointing out their agents for this site as a reference. Adwords for tooling!!" — geekymartian
• It's a category, not a one-off: commenters point to Stack Overflow's already-launched agents.stackoverflow.com as prior art for an "agentic Stack Overflow," and the founder's stated thesis is that blog-post AEO "will only work until agents stop using the web like humans do. This is a first step in that direction."

Why it matters for the market map

The brief's "who do they sell to" question now has a new intermediary to track: when agents do the choosing, vendors will optimize for Agent Experience (AX) the way they optimized for SEO/AEO — and whoever aggregates agent-written experience data (agent.reviews, Armature, Stack Overflow for agents, the GEO/visibility tools already spawning on HN like CiteRank) sits on the research-tool market's discovery chokepoint. For Hey, Lefty this cuts both ways: its own recommendability inside harness ecosystems (Claude Code, Codex, etc.) becomes a distribution channel it must manage; and as an orchestrator, agent-written review corpora are a candidate input source — with exactly the provenance and gaming caveats above.