What the evidence supports
- All five selected homepages returned a successful response to the disclosed research user agent; four exposed a canonical tag in the fetched HTML.
- Four of five exposed Organization schema and two exposed Product, SoftwareApplication or WebApplication schema on the homepage.
- One of five successful sites explicitly named at least one tracked AI crawler in robots.txt, while four served a valid markdown-like llms.txt file.
- These signals describe machine-readable implementation. They do not show whether a tool earns more citations, trials, upgrades or retention.
What the AI prosumer sample showed
Presence checks were regex-based against fetched public responses. Raw records, statuses and SHA-256 hashes are published so the snapshot can be inspected.
| Observed signal | Count | Practical interpretation |
|---|---|---|
| Successful homepage | 5 of 5 | Every selected product returned a usable public response in this run. |
| Canonical | 4 of 5 | The unresolved site needs rendered inspection before a definitive missing-canonical claim. |
| Organization schema | 4 of 5 | Entity markup was common in this small selected cohort. |
| Product or app schema | 2 of 5 | Homepage absence does not rule out supported markup on product-specific URLs. |
| Valid llms.txt | 4 of 5 | Adoption was high in this sample, but the file is optional and is not a citation guarantee. |
A stronger AI product discovery system
Document the product truth
Keep model capabilities, inputs, outputs, platform support, privacy, limits and pricing consistent across product, help and comparison pages.
Show verifiable use cases
Publish reproducible examples, evaluation methods and output boundaries for the jobs prospective users actually compare.
Measure discovery to activation
Track non-branded search, AI citations, qualified product visits, activation and paid conversion as related but separate stages.
Methodology
- Selected five recognizable self-serve AI-enabled prosumer products before collection: Otter, Descript, Framer, Luma and ElevenLabs.
- Requested each public homepage, robots.txt and llms.txt with a disclosed RadiusRankResearchBot user agent and a 12-second timeout.
- Recorded final URL, status, canonical presence, JSON-LD type presence, robots signals, explicit AI crawler agents and valid llms.txt presence.
- Saved response hashes and kept every selected site in the raw data rather than removing inconvenient observations.
Limitations
- The five-site purposive sample is not statistically representative of AI products or prosumer software.
- Regex checks do not validate schema accuracy, rendered content, product quality, usability, rankings or commercial outcomes.
- A public response can vary by location, user agent, cookies, JavaScript and edge configuration.
- llms.txt is optional and its presence does not prove that an answer engine used or cited the file.
Questions
Does an AI product need llms.txt to appear in AI answers?
No. A valid llms.txt file can provide a curated reference, but crawlable product pages, accurate documentation, standard access controls and independent evidence remain more important foundations.
Does product schema guarantee more trials or AI citations?
No. Supported structured data can clarify machine-readable facts, but it does not guarantee rankings, citations, traffic, activation or revenue.
Sources and data
- RadiusRank five-site AI prosumer readiness sample: Public homepage, robots.txt and llms.txt observations for Otter, Descript, Framer, Luma and ElevenLabs, including response hashes and limitations.
- Google Organization structured data: Official guidance for organization identity and administrative details.
- Google Software app structured data: Official software application markup guidance and eligibility details.
Found a changed price, product limit or source? Email team@radiusrank.com with the official URL. We preserve the snapshot date instead of silently rewriting historical data.