citare
GEO pillar — diagnostics

The four-index reality of AI search

Same brand. Same query. Four different answers. AI search isn't one surface — it's four indexes powering five platforms, each with its own crawler, scoring function, and audience. Ranking strongly in one tells you nothing about the others.

Updated May 2026

TL;DR

  • 1.Five AI search platforms run on four different indexes. AIO and Gemini share Google's. ChatGPT uses Bing's. Claude uses Brave's. Perplexity built its own.
  • 2.Surface rates across platforms regularly differ by 30-40 percentage points for the same brand on the same week. The variance is structural, not measurement noise.
  • 3.Five diagnostic archetypes recur. The shape of the gap between platforms tells you which fix to apply — Bing infrastructure, entity graph, foundation health, or platform-specific work.
  • 4.The useful metric is a per-platform breakdown, not an aggregate "AI visibility" score. Aggregates hide the actionable signal.

Five surfaces, four indexes, four crawlers

Each AI platform answers two architectural questions: what data does the model see (the grounding source), and how does it decide what to surface (the ranking logic). Grounding source is where the four-index reality starts.

SurfaceGrounding indexCrawler(s)
Google AI OverviewGoogle indexGooglebot + Google-Extended
GeminiGoogle index (different routing)Googlebot + Google-Extended
ChatGPT (web search)Bing indexBingbot (search) + GPTBot (training) + OAI-SearchBot
Claude (web search)Brave indexClaudeBot
PerplexityOwn index + multi-sourcePerplexityBot

Google AI Overview

Google index

The AI block above organic results. Geo-aware — same query, different cities, different brands. Cites pages that don't rank top-10 organically 62% of the time.

Gemini

Google index (different routing)

Shares Google's index with AIO but weights Knowledge Graph entity signals more. Surfaces named-competitor comparisons aggressively. Workspace integration matters.

ChatGPT (web search)

Bing index

Grounds via Bing, not Google. A brand at #1 on Google can be invisible to ChatGPT if Bing's coverage is thin. Most common gap in Indian brands and US SMBs.

Claude (web search)

Brave index

Anthropic's Claude grounds via Brave Search. Independent of Google and Bing. Thinner coverage but growing — meaningful in technical and B2B audiences.

Perplexity

Own index + multi-source

Runs its own crawler and builds its own index. Cites 5-10 sources per answer with linkable URLs — the most directly measurable AI platform. Rewards original research above all else.

Why the asymmetry is structural, not random

The mental model most marketing teams import from SEO is wrong for AI search. In SEO, you optimize once for Google and the smaller engines follow similar enough principles that good Google SEO translates. Optimization effort is roughly fungible.

In AI search this fungibility breaks. Each platform's architectural choices propagate through the entire experience — what gets crawled, what gets cited, how often the index updates, what content format the model prefers. A page perfectly optimized for AIO can be invisible on ChatGPT because ChatGPT's web search uses Bing, and Bing might not have crawled the page yet.

The four scoring functions in plain terms:

  • AIO — Google's index, weights structured data, FAQ schema, freshness, semantic completeness, geo-context.
  • Gemini — Google's index again, but weights Knowledge Graph entity strength, comparison content, conversational coherence.
  • ChatGPT — Bing's index plus OpenAI's offline training data, weights Bing rank, third-party source mentions, content depth.
  • Perplexity — its own index, weights original research, source-quality content, recent dateModified, backlinks (more than other platforms do).

Same content, four different scoring runs, four different surface rates. The variance is the system working as designed.

Five diagnostic archetypes

In our brand audits, five patterns recur often enough to be diagnostic. Each has a specific cause and a specific fix. Read your per-platform breakdown the way a doctor reads a blood panel — the shape of the gap is the signal.

1

Strong AIO, weak ChatGPT

Signal: AIO surface rate above 15%, ChatGPT under 10%, gap of 2x or more.

Cause: Bing index gap. AIO sources from Google, ChatGPT sources from Bing. Your Google indexing is fine; Bing's coverage of your domain isn't.

Fix:

  • Submit sitemap to Bing Webmaster Tools
  • Verify Bingbot allowed in robots.txt
  • Manually request indexing on top 20 priority pages through BWT
  • Audit Bing index coverage monthly

Timeline: 4-8 weeks to see ChatGPT surface rate move from sub-10% to 15-25%.

2

Strong AIO, weak Gemini

Signal: Gemini surface rate is 50% or less of AIO rate even though both source from Google.

Cause: Entity graph weakness. Gemini weights Knowledge Graph signals more heavily than AIO. No Wikipedia entry, sparse sameAs, no claimed knowledge panel — Gemini cites you less reliably.

Fix:

  • Expand Organization JSON-LD sameAs array (LinkedIn, Crunchbase, social channels, press coverage)
  • Pursue Wikipedia presence where editorially warranted
  • Claim and complete your Google knowledge panel
  • Consistent NAP across every directory

Timeline: 60-120 days as Knowledge Graph signals propagate.

3

Strong Perplexity, weak across the other three

Signal: Perplexity surface rate above 20%, AIO/Gemini/ChatGPT all under 15%.

Cause: Research-led content without foundation health. Perplexity rewards depth and original data above index size; the other three need basic crawlability, schema, and entity work.

Fix:

  • Audit robots.txt — confirm GPTBot, Bingbot, Googlebot, Google-Extended all allowed
  • Add Organization + FAQPage + Product schema to priority pages
  • Fix JavaScript-rendered content (the most common silent kill)
  • Submit to Bing Webmaster Tools

Timeline: 30-90 days. The fixes are foundational, not content-dependent.

4

Weak across all four

Signal: No platform crosses 10% surface rate.

Cause: Foundation issues. Either robots.txt blocks AI crawlers, critical content is in JavaScript or images, or the brand is too new for any platform to have indexed.

Fix:

  • Diagnostic order: robots.txt → render health → entity graph → content depth
  • Run an AI crawler access audit (see our /tools/ai-robots-checker)
  • Render core content in HTML, not JS-only
  • Build foundational pages (homepage, products, about) before chasing platform-specific tactics

Timeline: 90-180 days. Foundation work compounds — the fixes flow through to all four platforms simultaneously.

5

Strong on three, weak on one specific platform

Signal: One platform is conspicuously below the others by 30%+ despite the rest being healthy.

Cause: Platform-specific issue. Most often: one crawler blocked, one content type missing (e.g., no Knowledge Graph for Gemini, no original research for Perplexity, no FAQ schema for AIO).

Fix:

  • Diagnose against the platform's grounding source and ranking signals
  • AIO: add FAQ schema, freshness signals, structured H2/H3
  • Gemini: Knowledge Graph + comparison content
  • ChatGPT: Bing index health
  • Perplexity: original research, backlinks, source-quality content

Timeline: Platform-specific — usually 30-90 days.

The measurement implication

The most important consequence of the four-index reality: you cannot generalize across platforms. Aggregate "AI visibility" is a weak metric. The useful metric is a per-platform breakdown combined with citation context per mention.

This is also why traditional SEO platforms don't produce useful AI search measurement. Tools built around Google rank tracking have no native architecture for dispatching real queries against five platforms with persona context. The data model is different, the dispatch infrastructure is different, the metrics are different.

A proper AI search visibility tracker does five things at once:

  1. Dispatches real queries to all five platforms, not proxy signals
  2. Runs queries with multiple personas to capture variance
  3. Parses platform-specific response formats (AIO HTML block, ChatGPT chat output, Gemini chat output, Claude chat output, Perplexity citation list)
  4. Computes per-platform surface rate and citation context
  5. Benchmarks against named competitors in the same category

The five-platform optimization checklist

A practical sequence for moving from archetype 1 or 2 toward balanced five-platform visibility. Foundational work first, then platform- specific.

Crawler access

  • Allow GPTBot, ClaudeBot, PerplexityBot, Google-Extended in robots.txt
  • Verify each with its user-agent test
  • Audit /tools/ai-robots-checker monthly

Bing infrastructure

  • Submit sitemap to Bing Webmaster Tools
  • Verify domain ownership
  • Manually index top 20 priority pages
  • Monitor Bing index coverage monthly

Structured data

  • Organization schema on homepage with full sameAs array
  • LocalBusiness schema if physical presence
  • Product / FAQPage / Article / HowTo on relevant pages

Entity graph

  • Wikipedia presence where editorially appropriate
  • LinkedIn, Crunchbase, social profiles
  • Claim Google knowledge panel
  • Consistent NAP across directories

Content depth

  • Original research and data
  • Pillar + cluster topic architecture
  • Named-comparison content
  • Refresh cadence with dateModified

Measurement loop

  • Per-platform surface rate tracked weekly
  • Persona-anchored query dispatch
  • Named-competitor benchmarking
  • Citation context attribution

Frequently asked questions

What does 'four-index reality' mean?

Four major AI search platforms (Google AI Overview, Gemini, ChatGPT, Perplexity) ground their answers in four different indexes — Google's, Google's again with different routing, Bing's, and Perplexity's own. A fifth platform (Claude) grounds in a fifth index (Brave). One brand can rank strongly in some indexes and not in others, which is why optimizing for AI search isn't a single project.

Why does ChatGPT not see my brand if I rank #1 on Google?

ChatGPT's web search grounds against Bing's index, not Google's. A brand can be #1 on Google and missing from Bing entirely. The fix is Bing Webmaster Tools submission, ensuring Bingbot is allowed in robots.txt, and verifying Bing has crawled your priority pages. This is the single most common gap in Indian brands and US SMBs that ignored Bing for a decade.

Is the AIO citation just the top organic results?

No. 62% of pages cited in Google AI Overviews do not rank in the top 10 organically for the same query. AIO uses Google's index but applies a separate relevance and synthesis evaluation that weights FAQ schema, semantic completeness, freshness, and structured data more than rank position.

Does Perplexity use Google's index?

No. Perplexity runs its own crawler (PerplexityBot) and builds its own index, independent of both Google and Bing. A brand can be invisible across Google's surfaces and still appear on Perplexity if PerplexityBot has crawled the site and the content matches Perplexity's source-quality criteria.

How is Gemini different from Google AI Overview?

Both source from Google's index but they are distinct products with different scoring. AIO is the answer block above the search results page. Gemini is Google's standalone chat assistant. Gemini surfaces named-competitor comparisons more aggressively, weights Knowledge Graph entity signals more heavily, and retains conversational state across turns. The same brand can have a 30% AIO surface rate and a 10% Gemini surface rate (or vice versa).

Should I optimize for all five platforms or focus on the strongest one?

Foundational work — crawlability, structured data, content depth, entity graph — helps all five simultaneously. Platform-specific work (Bing for ChatGPT, Knowledge Graph for Gemini, original research for Perplexity, FAQ schema for AIO) compounds when done in parallel. Don't pick one platform; sequence the foundational fixes first, then layer platform-specific work.

How long does it take for AI platforms to learn about a new brand?

Web-search modes (AIO, Gemini, ChatGPT web search, Perplexity) update within days to weeks of new content. Trained-knowledge modes (the non-web-search portion of base ChatGPT) lag by full training cycles — often 6-12 months. New brands should expect 4-8 weeks for web-search visibility and longer for trained-knowledge inclusion.

How do I read my per-platform breakdown?

Look at the gap between platforms, not just the absolute numbers. AIO vs ChatGPT measures your Google-vs-Bing index health. AIO vs Gemini measures your Knowledge Graph entity strength. Perplexity vs the others measures your content depth and backlink authority. The shape of the gap is the diagnostic.

See your surface rate across all five platforms

Citare's Brand Radar runs persona-anchored query dispatches against Google AIO, Gemini, ChatGPT, Claude, and Perplexity. Get a per-platform breakdown, named-competitor benchmark, and the diagnostic that points to your specific fix.

Related