- Visibility21
- Citations1
- Sentiment65
- Site quality87
- AI readiness21
AI assistants name OpenAI in 16.7% of answers, more often than the typical company in the technology sector. Its overall score is 36 out of 100, ranking 731st of 2,047 companies. The company's website passes 3 out of 14 technical checks of how ready it is for AI tools.
We asked 15 unbranded questions about the technology market: none named OpenAI. AI assistants named it in 10 of 60 answers (16.7%). The typical technology company is named in 10.0% of answers. By how often it is named, OpenAI ranks 98 of 273 in its sector.
Every dot is a company in the study. Hover one to see it; click to open its report.
This companySame sectorEvery other companyNever named
Five parts, each scored out of 100.
How often AI assistants name the company
How often AI uses the company's own website as a source
How positively AI describes the company
How well built the website's own pages are
How easily AI tools can read and use the website
Visibility is calculated from the answers on this page. How the score is built.
273 companies in this sector. The line on each bar is the typical company.
An answer that names the company
“Alternatives to large-scale AI research platforms for startups and small teams”
Good question — startups and small teams don’t need to (and often shouldn’t) try to replicate the big, expensive AI research platforms. Here’s a practical guide to alternative approaches, tools, vendors and cost-saving patterns you can use to get results without massive infrastructure or headcount.
High-level approaches
Build with managed cloud ML services: use inference/hosting, pipelines and training-as-a-service instead of owning infra. This cuts ops work and scales with demand.
Use open-source models and local/colocated GPUs: run and fine-tune smaller or community models on modest GPU …
The AI’s answer as we received it, shortened and not checked for accuracy.
How the mentions in these answers were split between this company and its rivals
Answers naming it, out of answers received
What this website publishes for an AI agent to read, use and buy from, on a scale of six levels.
- 0
- 1
- 2
- 3
- 4
- 5
Level 1: Basic web presence. Two of robots.txt, a sitemap and Link headers are in place.
- Content Signals
The website turns away requests that do not come from a browser. AI agents that say what they are meet the same refusal.
Show every check, for your technical teamHide the checks
- study: 92.0%robots.txtPassed
A valid robots.txt with crawl rules at the site root.
- study: 86.8%SitemapPassed
An XML sitemap listing the pages, ideally referenced from robots.txt.
- study: 0.3%Link headersNot met
Send Link headers such as rel="api-catalog" or rel="describedby" on the homepage.
- study: 5.5%DNS for AI DiscoveryNot present
Advertise agent endpoints with SVCB records under _agents.
- study: 1.9%Markdown negotiationNot met
Return a markdown version when asked with Accept: text/markdown, for example at the CDN.
- study: 25.1%llms.txtNot met
Publish an llms.txt with a short summary and the key links.
- study: 93.4%AI bot rulesPartly met
Add explicit robots.txt groups for the main AI crawlers.
- study: 1.4%Content SignalsNot met
Add a Content-Signal line for search, ai-input and ai-train to robots.txt.
- study: 0.2%Web Bot AuthNot present
Only relevant if you run your own agents or crawlers: publish a signing key directory.
- study: 0.3%API CatalogNot met
List public APIs in a linkset at /.well-known/api-catalog.
- study: 11.8%OAuth discoveryNot met
Publish OpenID Connect or OAuth server metadata under /.well-known.
- study: 10.3%OAuth Protected ResourceNot met
Serve protected resource metadata at /.well-known/oauth-protected-resource.
- study: 0.0%auth.mdNot met
Publish an auth.md describing how agents sign in.
- study: 0.2%MCP Server CardNot met
Offer an MCP server and publish its server card under /.well-known.
- study: 0.1%A2A Agent CardNot met
Publish an agent card at /.well-known/agent-card.json.
- study: 0.2%Agent SkillsNot met
Publish a skills index with the main tasks agents can do.
- study: 33.7%WebMCPNot met
Register key actions such as search or cart as WebMCP tools.
- study: 0.1%ARD manifestNot met
Publish an ai-catalog.json listing every agent interface.
Score 24 of 100 · 100% of checks could run · checked on October 3, 2026 · openai.com · 2,047 companies scanned
3 of 14 passed. This list is for your technical team. The percentage is how many companies in the study pass each check.
These 14 checks ran during the audit and are what the AI readiness part of the overall score counts. The scan above ran later and is stricter on Link headers and Markdown, so a check can pass here and not there.
Show the 14 checksHide them
- Passedrobots.txt publishedstudy: 85.0%
- PassedXML sitemapstudy: 73.2%
- Not metHTTP Link headers (RFC 8288)study: 29.0%
- Not metMarkdown content negotiationstudy: 8.3%
- Not metllms.txt publishedstudy: 21.7%
- PassedToken budget (page weight)study: 95.2%
- Not metExplicit AI bot rulesstudy: 13.0%
- Not metAPI Catalog (.well-known/api-catalog)study: 0.2%
- Not metOAuth Authorization Server discoverystudy: 10.1%
- Not metOAuth Protected Resource discoverystudy: 9.3%
- Not metMCP Server Cardstudy: 0.0%
- Not metA2A Agent Cardstudy: 0.0%
- Not metWebMCP toolsstudy: 4.2%
- Not metAgent Skills declaredstudy: 0.0%
This website sets no rules for any AI crawler by name, like 84% of the websites whose rules we could read. Every AI crawler follows its general rules.
- GPTBotpartly
- ClaudeBotpartly
- Anthropic-AIpartly
- Google-Extendedpartly
- Applebot-Extendedpartly
- Meta-ExternalAgentpartly
- FacebookBotpartly
- Bytespiderpartly
- CCBotpartly
- Diffbotpartly
- Omgilibotpartly
- OAI-SearchBotpartly
- Claude-SearchBotpartly
- PerplexityBotpartly
- Applebotpartly
- Amazonbotpartly
- YouBotpartly
- ChatGPT-Userpartly
- Claude-Userpartly
- Perplexity-Userpartly
- DuckAssistBotpartly
- MistralAI-Userpartly
Read from the website’s own rules file (robots.txt). An asterisk marks a crawler the file names. The rest follow its general rules. “Partly” means only some pages are blocked.
How many times each website was listed as a source in these answers
In the study as Large private companies.
- How does an AI software platform compare to building custom AI solutions from scratch?
- What factors affect the cost of implementing advanced AI tools for developers?
- Alternatives to large-scale AI research platforms for startups and small teams
- Is investing in an enterprise AI software platform worth it for small businesses?
- Do advanced conversational AI tools justify their cost for solo developers?
- Other options besides cloud-based AI platforms for experimenting with language models
- What should I do if an AI research tool is returning unexpected results?
- Top artificial intelligence tools for individual users interested in exploring language models
- Why might an AI-powered language model misunderstand user input?
- Ways to leverage AI research software for automating customer support tasks
- What are the best AI research platforms for businesses looking to integrate advanced conversational tools?
- Best entry points for developers wanting to get started with advanced AI solutions
- Typical pricing tiers for enterprise-level artificial intelligence software platforms
- How can teams use AI-driven content generation tools to streamline business operations?
- Benefits of general-purpose AI platforms versus specialized machine learning tools
Columns, left to right: OpenAI, Claude, Perplexity, Gemini. The AI assistants often disagree. Across the study, on 40% of the questions where a company is named, only one of the four names it.
A badge of this result that you can show on your own website. Copy the code below.
<a href="https://aeo-audit.rezolve.com/us/openai.com"><img src="https://aeo-audit.rezolve.com/us/badges/openai.com.svg" alt="Overall score 36 of 100, US AI Visibility Study 2026" width="280" height="64"></a>