- Visibility3
- Citations0
- Sentiment50
- Site quality80
- AI readiness50
AI assistants rarely name Innodata: it appears in 1 of 60 answers. For comparison, Labelbox, Scale AI, and Appen are also named in that answer. Innodata's overall score is 29 out of 100, ranking 1,341st of 2,047, which is lower than the typical company in its sector. Its website passes 7 of 14 technical checks for AI tool readiness.
We asked 15 unbranded questions about the technology market: none named Innodata. AI assistants named it in 1 of 60 answers (1.7%). The typical technology company is named in 10.0% of answers. By how often it is named, Innodata ranks 211 of 273 in its sector.
Named fewer than three times in total. Read this as a sign that AI assistants rarely bring this company up, not as an exact comparison with its rivals.
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
“Top services for supervised fine-tuning of generative AI models in an enterprise setting”
Based on current enterprise offerings, here are the top services for supervised fine-tuning of generative AI models:
Cloud Provider Platforms
Google (Gemini via Vertex AI) and Anthropic (Claude via Amazon Bedrock) remain the primary closed-source options for new fine-tuning workloads. Vertex AI supports supervised fine-tuning where developers train models with labeled examples so they learn to produce desired outputs for specific tasks. Amazon Bedrock includes reinforcement fine-tuning capabilities that allow developers to customize models based on feedback signals.
Tuning-as-a-Servic …
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
Not by itself. Innodata publishes none of the purchase protocols we look for. An agent would have to click through the checkout like a person.
- UCPNot present
- ACPNot present
- AP2Not present
- x402Not present
- MPPNot present
The shop runs on WooCommerce.
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 rulesPassed
Explicit robots.txt rules for AI crawlers such as GPTBot, ClaudeBot and PerplexityBot.
- 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 discoveryPartly met
Publish OpenID Connect or OAuth server metadata under /.well-known.
- study: 10.3%OAuth Protected ResourcePassed
Metadata telling agents which authorization server to use.
- 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%WebMCPCould not check
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 38 of 100 · 97% of checks could run · checked on October 5, 2026 · innodata.com · 2,047 companies scanned
7 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%
- PassedHTTP 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%
- PassedExplicit AI bot rulesstudy: 13.0%
- Not metAPI Catalog (.well-known/api-catalog)study: 0.2%
- PassedOAuth Authorization Server discoverystudy: 10.1%
- PassedOAuth 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 rules for 10 AI crawlers by name. All others follow its general rules.
- GPTBot*allowed
- ClaudeBot*allowed
- Anthropic-AIpartly
- Google-Extended*allowed
- Applebot-Extendedpartly
- Meta-ExternalAgent*allowed
- FacebookBotpartly
- Bytespiderpartly
- CCBotpartly
- Diffbotpartly
- Omgilibotpartly
- OAI-SearchBot*allowed
- Claude-SearchBot*allowed
- PerplexityBot*allowed
- Applebotpartly
- Amazonbotpartly
- YouBotpartly
- ChatGPT-User*allowed
- Claude-User*allowed
- Perplexity-User*allowed
- 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 US-listed.
- What are some alternatives to classic red teaming for evaluating generative AI risks?
- Alternatives to full-service AI data annotation for companies with in-house data teams
- What factors influence the cost of supervised fine-tuning for generative AI models?
- What role does red teaming play in testing AI systems for enterprise security and compliance?
- Is it worth investing in external AI data collection for enterprise machine learning projects?
- Best US-based AI consulting firms specializing in large-scale enterprise model deployment
- How can structured data collection improve the accuracy of generative AI deployments in enterprises?
- Best ways to overcome challenges with fine-tuning large language models using proprietary datasets
- AI data red teaming vs model evaluation: which is more effective for risk mitigation?
- Are comprehensive AI consulting packages a good fit for scaling generative AI in a large organization?
- What are the best enterprise AI data solution providers for large organizations?
- How are enterprise AI data solution services typically priced in the US?
- How to address inconsistent annotation quality in outsourced AI data labeling projects?
- Top services for supervised fine-tuning of generative AI models in an enterprise setting
- Difference between managed AI data annotation services and self-serve annotation platforms
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.