How to optimize for Google AI features, ChatGPT Search, Claude, and answer engines without platform myths

By licheo Research Team | | GEO Guides | 16 min read

This guide draws a hard line between documented controls and reasonable inference.

That distinction matters. In traditional SEO, people already overstate certainty. In GEO, the problem is worse because platforms change quickly and often publish less operational detail than SEOs would like.

So the rule for this guide is simple:

  • If a platform documents a control, we treat it as a control.
  • If a platform does not document a ranking factor, we do not pretend it did.

If you want the foundational version first, start with our complete GEO guide.

The universal layer comes first

Before you split by platform, fix the shared layer:

  • crawlability
  • indexability
  • clear structure
  • explicit entities
  • good sourcing
  • content that is actually worth citing

That is the baseline for every platform discussed below.

Google AI Overviews and AI Mode

Google is the platform with the clearest public documentation in this area, and the main takeaway is more conservative than most GEO hot takes:

Google says standard SEO best practices remain relevant, and there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode.

Source:

What you can control on Google

1. Search eligibility

Google says a page must be:

  • indexed
  • eligible to appear in Google Search
  • eligible to show a snippet

If a page is not indexable or is blocked from snippets, your GEO discussion should stop there and move into technical triage.

2. Preview controls

Google specifically points site owners to these controls:

  • nosnippet
  • data-nosnippet
  • max-snippet
  • noindex

Use them intentionally. Restricting previews can limit how much Google can display from the page in Search experiences.

3. Content quality

Google points back to:

  • helpful, reliable, people-first content
  • page experience
  • text availability
  • strong internal links
  • structured data that matches visible text
  • current Merchant Center and Business Profile information where relevant

That is a stronger and more useful checklist than any speculative "AI overview hack."

4. Structured data

Google explicitly says you do not need special AI files or special schema to appear in AI features. But structured data still helps Google understand page content.

The right interpretation is:

  • use schema for clarity
  • do not use schema as superstition

5. Google-Extended

Google's crawler documentation says Google-Extended is a product token publishers can use to manage some Gemini training and grounding uses. It also says Google-Extended does not affect inclusion in Google Search and is not used as a ranking signal.

That means:

  • block or allow Google-Extended based on your training and grounding preferences
  • do not confuse it with Google Search visibility

Sources:

What not to assume on Google

Do not assume:

  • a page needs special AI markup
  • Google-Extended is required for AI Overviews
  • getting indexed means getting surfaced
  • adding FAQ schema guarantees inclusion

Google does not document any of those claims.

ChatGPT Search

OpenAI's ChatGPT Search documentation gives site owners three especially useful facts:

  1. ChatGPT Search may rewrite user queries into one or more targeted searches.
  2. Search responses include inline citations and a source panel.
  3. To make sure your site is available in ChatGPT Search, it is important to allow OAI-Searchbot and ensure the host or CDN allows traffic from OpenAI's published IP ranges.

Sources:

What you can control for ChatGPT Search

1. Bot access

If OAI-Searchbot cannot crawl your site, inclusion becomes much harder. Review:

  • robots.txt
  • WAF rules
  • CDN bot management
  • origin allowlists

An example starting point:

User-agent: OAI-Searchbot
Allow: /

That alone is not enough, but without it you may be losing before content quality even enters the conversation.

2. Source suitability

Because ChatGPT Search shows inline citations, the page benefits from having:

  • a direct answer near the top
  • sections that can stand on their own
  • visible dates where freshness matters
  • named sources
  • explicit definitions and comparisons

This is partly documented and partly inference. The documented part is that ChatGPT Search cites sources. The inferred part is that a page is easier to cite when its best statements are easy to lift accurately.

3. Local and commercial specificity

OpenAI documents that ChatGPT Search may use general location information and, when enabled, device location to improve local results. It also documents shopping behavior that depends on structured metadata from first- and third-party providers.

That means local and commercial pages should be especially careful about:

  • keeping business details current
  • keeping product and merchant data accurate
  • making location and service area explicit

Sources:

What not to assume for ChatGPT Search

Do not assume:

  • a single crawlable page guarantees inclusion
  • ChatGPT Search uses the same weighting as Google Search
  • brand size alone will compensate for weak pages
  • ranking tools designed for blue links tell you enough about citation behavior

OpenAI does not publish a public ranking formula for citations, so you should not pretend otherwise.

Claude and Anthropic

Anthropic documents its bots more clearly than most companies. That makes Claude-related optimization cleaner to reason about.

Anthropic distinguishes among:

  • ClaudeBot for model development
  • Claude-User for user-directed retrieval
  • Claude-SearchBot for search-result quality

Anthropic also says its bots respect robots directives and support Crawl-delay.

Source:

What you can control for Claude

1. Intent-based bot access

You can choose whether to allow:

  • model-development crawling
  • user-requested retrieval
  • search-quality crawling

That is more granular than many teams realize.

For example:

User-agent: ClaudeBot
Disallow: /

User-agent: Claude-User
Allow: /

User-agent: Claude-SearchBot
Allow: /

That is not a universal recommendation. It is an example of how to think in terms of bot intent rather than "allow Anthropic" as one indivisible choice.

2. Content layout

Because Anthropic's retrieval-related bots are still trying to fetch and use public web content, the same universal GEO priorities apply:

  • strong summaries
  • clean hierarchy
  • source-rich claims
  • readable prose
  • predictable page structure

This is partly documented and partly inferred. Anthropic documents the bot roles. It does not publish a detailed citation-scoring formula.

What not to assume for Claude

Do not assume:

  • blocking ClaudeBot also blocks user-directed retrieval
  • all Anthropic traffic has the same purpose
  • IP allowlists are a stable long-term solution, since Anthropic says it does not currently publish fixed IP ranges

Engines with limited public documentation

Some answer engines publish less operational guidance than Google, OpenAI, or Anthropic. In those cases, the safest play is to optimize the universal layer rather than chase undocumented platform-specific lore.

That means:

  • keep pages accessible and linkable
  • use clean headings
  • publish clear summaries
  • include references
  • keep pages updated
  • surface author and organization context
  • submit changed URLs quickly where supported

This is where IndexNow becomes useful. It is not an AI-specific protocol, but it helps participating engines learn that a URL changed without waiting for ordinary crawl discovery.

The platform control matrix

Here is the plain-English version.

For Google AI features

Prioritize:

  • indexability
  • snippet eligibility
  • structured data accuracy
  • people-first content
  • internal linking

Ignore:

  • special AI markup myths
  • claims that Google-Extended controls AI Overviews

For ChatGPT Search

Prioritize:

  • OAI-Searchbot access
  • CDN and WAF allow rules
  • clear source blocks
  • answer-first layouts
  • commercial and local data completeness

Ignore:

  • unsupported claims about exact citation weighting

For Claude

Prioritize:

  • intentional bot policy
  • clarity about which Anthropic bots you allow
  • strong text structure and sourcing

Ignore:

  • assumptions that all Anthropic traffic serves one purpose

For participating answer engines more broadly

Prioritize:

  • rapid URL change submission
  • stable public HTML
  • content worth citing

Ignore:

  • the belief that every engine needs a separate editorial strategy on day one

The real workflow

In practice, the best teams do platform optimization in this order:

  1. fix technical blockers
  2. rebuild a small set of priority pages
  3. make bot access intentional
  4. publish supporting content that clarifies entities and subtopics
  5. track which pages are actually cited or surfaced
  6. iterate from evidence, not from theory

That is slower than chasing every new rumor, but it is how you avoid building a GEO program on invented facts.

Final takeaway

Platform-specific GEO is mostly about knowing which controls are real.

Google gives you eligibility and preview rules. OpenAI gives you Searchbot access rules. Anthropic gives you role-specific bot controls. IndexNow gives you a speed layer for change discovery.

Everything else still comes back to content quality, structure, evidence, and trust.

Sources and further reading