2027 CONCEPT
Where do you compete?
Platform-specific controls

One business. Different discovery systems.

Work on the shared foundations, then check each product’s documented access controls and observe its actual answers.

Documented

Access controls

Provider documentation explains which bots and settings govern particular uses.

Practical recommendation

Evidence & clarity

Publish enough truthful detail for a buyer to evaluate your business and inspect supporting evidence.

Must be observed

Answer selection

Test whether the business is actually mentioned, cited or recommended for specific questions.

G

Google Search

Classic results · AI Overviews · AI Mode

Documented access

Check Googlebot access, indexability and snippet eligibility. In Search Console, review Settings → Search generative AI, including inherited settings. The documented default is inclusion.

What to do

Maintain useful pages and accurate business/product data. Inspect ordinary search performance alongside the Generative AI performance report. Eligibility does not guarantee a link or recommendation.

Interpretation limit: Google’s AI optimization guide says llms.txt and special AI markup are not required. A knowledge panel is not a universal entry condition.

G

Gemini apps

A distinct product and control surface

Documented access

Google-Extended controls specified uses of crawled content for Gemini training and grounding. It has no separate HTTP crawler identity and does not determine Google Search ranking.

What to do

Review Google-Extended deliberately if Gemini-related uses matter to your business. Evaluate Gemini answers separately from Google AI Mode; do not assume identical citations or recommendations.

Interpretation limit: The public documentation does not expose a business recommendation formula. Access and availability are controllable; selection remains platform-specific.

O

ChatGPT

Search results · pages opened for a user

Documented access

OAI-SearchBot is the search crawler. GPTBot concerns model training. ChatGPT-User supports some user-initiated visits and is not the control for search inclusion.

What to do

Permit the search access you intend, including the network edge, and inspect referral analytics. Keep offers, identity and evidence publicly understandable. Test realistic questions with search enabled.

Interpretation limit: Search access does not guarantee placement. Allowing model training is not required merely to permit the separate search crawler.

A

Claude

Search results · pages opened for a user

Documented access

Claude-SearchBot supports search. Claude-User retrieves content for user requests. ClaudeBot concerns model development. Anthropic documents separate controls for these agents.

What to do

Review the relevant robots directives and infrastructure restrictions. Make important information accessible and current, then retain actual answer and citation observations in the journal.

Interpretation limit: Anthropic does not publish a general business ranking formula. A crawl log proves access, not that Claude recommended the business.

Developer example: selected discovery rules

This is an illustrative excerpt for public pages, not a replacement for your existing robots.txt. Preserve private-path restrictions in every applicable group. A developer must merge and test the rules against the complete site policy and firewall.

User-agent: OAI-SearchBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Claude-User
Allow: /

Review training preferences separately. Robots.txt is not authentication and must never be used to protect private information. Inspect Google Search and Google-Extended settings according to their separate purposes.

What makes a business useful to recommend? Our practical framework is fit + clear facts + supporting evidence + availability. This is a planning framework, not a disclosed internal formula shared by Gemini, ChatGPT or Claude.