For website owners and product teams
See which competitors AI recommends.
See which companies appear for your customers’ questions. Inspect the sources and your own pages, then decide what to improve.
Sign in to prepare an evaluation. You choose when to run it.
Know who appears. See the returned recommendations.
Inspect the sources. Read the evidence behind an answer.
Review what to change. Check your pages before editing.
A clear next step for your website.
Start with a customer question. Finish with evidence your team can act on.
Ask a useful question
Save the questions people ask when choosing a product like yours. Start an evaluation when you’re ready.
Inspect the evidence
See the returned companies and sources. Inspect your own page to check whether the relevant information is already there.
Make a change. Check again.
Use page checks and repair suggestions to decide what to edit. After your team publishes, run a fresh evaluation and review the evidence.
Each result records one run. A different answer after an edit does not establish that the edit caused it.
Questions worth checking.
Software is one starting point. Shops, service businesses, and learning platforms have their own questions to answer.
Each search ranking belongs to its exact question. The separate page reports use the same HTML rubric across all five audiences.
- Software & work
- Which plan includes what I need?
- Shops & brands
- What are the shipping and return terms?
- Services & travel
- Where is this available, and how do I book?
- Learning
- What will I learn, and what are the prerequisites?
- Developer tools
- How do I get an integration working?
See who gets recommended.
Explore real search recommendations, with the returned order and sources intact.
For a small Node.js app with a PostgreSQL database, how do Vercel, Render, Railway and Cloudflare Workers compare on setup, recurring cost and runtime limits?
en-US · Language: en
- Render
render.com
1Returned reason
First choice for a conventional Node server and PostgreSQL, with a straightforward deployment model and a $13 monthly compute baseline before storage and usage.
- Railway
railway.com
2Returned reason
Flexible deployment and resource-based billing suit a small app, provided the combined app and database usage is budgeted beyond the $5 minimum.
- Vercel
vercel.com
3Returned reason
Good fit for framework-oriented, request-driven applications; external PostgreSQL costs and function duration limits make it less direct for a persistent Node server.
- Cloudflare Workers
developers.cloudflare.com
4Returned reason
Low entry compute cost for a lightweight API, with additional compatibility work, a 128 MB isolate limit, and separately hosted PostgreSQL.
About this observation
Position follows the returned list for this question. Depending on the question, recommendations may be websites, products, resources or steps. Another run can return a different order. Reasons and citations remain evidence to review.
- Model
- gpt-6-astra
- Observed at
- Surface
- OpenAI managed Agents API
- Search mode
- open-web
- Language / locale
- en / en-US
- Harness
- keyword-open-web-v2
- Environment
- openai_hosted
Sources returned with the observation
- Render pricing
- Render free service limits
- Railway pricing
- Railway networking specifications and limits
- Vercel pricing
- Vercel Functions limits
- Postgres on Vercel
- Cloudflare Workers pricing
- Cloudflare Workers limits
- Connect Workers to PostgreSQL
Limits of this result
- Recommendation order is an assessment for this query, not a measured search-engine rank, deployment benchmark, or independently verified performance comparison. Public discovery was not domain-restricted; the final comparison uses vendor documentation and is not exhaustive market coverage.
- No app code, traffic profile, memory measurements, database size, region, or backup requirements were supplied. Total costs, package compatibility, and production suitability were not tested; external PostgreSQL pricing was not compared.
- Render's pricing page omitted its pricing tables in the direct text read; its indexed official-page excerpts supplied the quoted prices. Opening Railway's Express guide returned an internal error; its search excerpt was available.
- Not every proxy timeout, quota, storage charge, or operational database responsibility was verified. Published limits and pricing can change; retrieved pages had different crawl dates.
HTML page checks for 36 websites
Open a saved page report to inspect technical findings alongside the search observations.
Folio page evaluation
Allbirds
2 checks need attention
1 failed checkScroll the table to see points and details.
| Check | Details | ||
|---|---|---|---|
| Page title | Pass | 15 / 15 | |
| Meta description | Pass | 10 / 10 | |
| Primary heading | Pass | 10 / 10 | |
| Canonical URL | Pass | 10 / 10 | |
| Indexing directive | Pass | 10 / 10 |
Captured source and hash
- Requested page
- https://www.allbirds.com/
- Captured page
- https://www.allbirds.com/
- Captured at
- 2026-09-13T04:44:54.033Z
- Capture SHA-256
f5129921fb34b5b075895e4e23657e97a8291cbd16ece6d776f93ec26e0161a9
The hash identifies captured text. Opening a source link visits the current website, which may have changed.
A few useful details.
What does a ranking mean?
It records a website’s position in one returned recommendation list for the displayed question, model, and date. Open the cited sources to inspect the answer. A later run can return a different list.
What does a page score tell me?
It records ten checks against captured HTML: titles, descriptions, headings, links, indexing instructions, readable text, and structured data. It does not measure search rank, product quality, or whether someone can finish a purchase.
How do the agent evaluations work?
A managed agent answers bounded questions from captured evidence. Folio checks its output and compares product and pricing answers with reference values you provide. Missing references stay visible, and each returned quote can be inspected in its source.
What happens after I find a problem?
Review technical repair suggestions and download the changes you choose. Once your team publishes a change, start a fresh evaluation and compare the saved results.
What will AI say about your website?
Choose the questions that matter to your customers.