Platforms We Analyze

AI Marketing Consensus Index studies recommendations produced by multiple leading artificial intelligence platforms. Our standard research universe currently includes:

• ChatGPT
• Gemini
• Claude
• Perplexity
• Grok
• DeepSeek
• Kimi

These platforms are analyzed independently. For standard consensus research: One platform = one vote. We do not give one AI system more influence because it has more users, greater market share, stronger brand recognition, or a larger parent company.

Why We Analyze Multiple AI Platforms

There is no single universal AI answer. A marketer asking:

Which GEO agencies would you recommend for a B2B SaaS company?

may receive different recommendations from:

• ChatGPT
• Gemini
• Claude
• Perplexity
• Grok
• DeepSeek
• Kimi

One platform may repeatedly recommend enterprise agencies. Another may emphasize software vendors. Another may surface smaller specialist firms. Another may rely heavily on recently published comparison pages. Those differences are exactly what AI Marketing Consensus Index is designed to measure.

One AI Answer Is Not Consensus

If ChatGPT recommends a company, that tells us something about ChatGPT's answer. It does not tell us: the AI market agrees. Consensus becomes more meaningful when the same company independently appears across several AI systems. For example: Company A — recommended by 6 of 7 platforms is a different signal from: Company B — recommended by 1 of 7 platforms Our research is designed to make that difference visible.

Our Current Standard Platform Set

ChatGPT

ChatGPT is an AI assistant developed by OpenAI. Depending on the available product experience, configuration, model, and research environment, ChatGPT may generate answers using:

• Model knowledge
• Search or browsing capabilities
• Retrieved web information
• Other available system tools

We record the platform as: ChatGPT and, where practical, separately record additional technical metadata about the model or mode used.

Gemini

Gemini is Google's generative AI platform. Gemini can produce answers using Google's AI systems and, depending on the environment and available functionality, may incorporate information from the web and other Google-connected sources. For standard consensus measurement, Gemini receives: one platform vote.

Claude

Claude is Anthropic's AI assistant. Claude may generate recommendations using model knowledge and, where available in the research environment, web or search functionality. Claude is treated as an independent platform in our standard consensus framework.

Perplexity

Perplexity is an AI answer and search platform with a strong emphasis on web-connected responses and visible source citations. Its recommendation behavior can differ significantly from conventional chatbot interfaces because current web information and cited sources can play a larger role in its answers. Perplexity receives one vote in standard research.

Grok

Grok is xAI's artificial intelligence platform. Depending on the available interface and research configuration, Grok may incorporate current web or platform-connected information when answering recommendation questions. Grok is treated as one independent platform.

DeepSeek

DeepSeek provides generative AI models and consumer-facing AI experiences. DeepSeek may produce recommendation sets that differ substantially from U.S.-based AI systems, making it useful for measuring broader cross-platform agreement and disagreement. DeepSeek receives one vote in standard research.

Kimi

Kimi is an AI platform developed by Moonshot AI. Including Kimi broadens the research beyond the most commonly discussed U.S.-based AI brands and can help reveal whether recommendation patterns persist across a wider set of AI ecosystems. Kimi receives one vote in standard research.

Why These Seven?

Our objective is not to claim that these are the only important AI systems. The current platform set is intended to provide a useful cross-section of prominent generative AI and AI-search environments. The set includes systems with different:

• Model families
• Product interfaces
• Search capabilities
• Retrieval systems
• Geographic origins
• Training histories
• Source-selection behavior
• Recommendation patterns

That diversity makes consensus more informative.

We Do Not Weight Platforms by Market Share

Suppose:

• ChatGPT recommends Company A
• Gemini recommends Company B
• Claude recommends Company B
• Perplexity recommends Company B
• Grok recommends Company B
• DeepSeek recommends Company B
• Kimi recommends Company B

We do not say: ChatGPT has more users, so its recommendation should count as four votes. The result is: Company A — 1 platform Company B — 6 platforms Our methodology measures: cross-platform recommendation agreement not: estimated consumer exposure weighted by platform usage.

One Provider Does Not Get Multiple Votes by Default

A single AI company may offer:

• Multiple models
• Multiple reasoning modes
• Multiple subscription tiers
• Search and non-search experiences
• API access
• Consumer applications

For standard consensus research, we generally treat the: platform as the voting unit. Running three different models from the same provider does not automatically create three independent platform votes.

Why We Use the Platform as the Stable Unit

Exact model names and versions can change quickly. A consumer may interact with: ChatGPT while the underlying model changes over time. The same is true for other platforms. Using the platform as the public research unit provides greater continuity. Where practical, we still preserve technical metadata such as:

• Model
• Mode
• Interface
• Search status
• Research date

This lets us maintain methodological detail without making the entire historical dataset dependent on rapidly changing model labels.

Model Information Still Matters

Using the platform as the primary vote does not mean model information is irrelevant. When practical, we record:

• Model identifier
• Mode
• Search / browsing state
• Interface type
• Date and time
• Other material research settings

These details can help explain changes in platform behavior.

Consumer Interfaces and APIs Are Not Necessarily Identical

An important limitation of AI research is that the same provider may return different answers through:

• Consumer web interface
• Mobile application
• API
• Search-enabled mode
• Non-search mode
• Different subscription tiers
• Different models

A ChatGPT API response should not automatically be assumed to reproduce exactly what a consumer sees in ChatGPT. Likewise for other platforms.

We Do Not Silently Substitute One Environment for Another

If research is intended to measure a particular platform experience, the research record should identify the actual environment used where practical. If a platform becomes unavailable and another interface is substituted, that should be documented rather than silently treated as equivalent.

Search and Web Access Matter

Some AI systems can access current web information during a response. Others may rely more heavily on model knowledge in a particular configuration. This can materially affect recommendations. A web-connected system may discover:

• A newly launched GEO agency
• A recently released AI visibility tool
• New pricing
• Current product features
• Recent comparison articles

while another system may surface older brands from its existing knowledge.

We Record Search State Where Practical

Where technically available, research metadata should identify whether the response involved:

• Web search
• Browsing
• Retrieval
• Search-disabled generation
• Unknown search state

This helps future interpretation of the dataset.

Search-Enabled and Non-Search Responses Are Not Automatically Equivalent

If one AI platform searches the current web and another does not, their responses may reflect different information environments. We do not hide this limitation. Cross-platform consensus measures the outputs produced by the included systems. It does not mean every system used identical information.

Citations Differ Across Platforms

Some AI platforms frequently provide visible citations. Others may:

• Cite only occasionally
• Provide links without formal citations
• Give no visible source information
• Use different source-selection systems

The absence of a citation does not automatically make a recommendation invalid. Likewise, the presence of a citation does not automatically make a recommendation factually correct.

Recommendation Data and Citation Data Are Separate

This distinction is central to AI Marketing Consensus Index. Suppose Perplexity: recommends Company A but cites: Company B's research article Then: Company A receives a recommendation. Company B receives a citation. Those are different research events.

A Citation Is Not an AI Vote

If an AI system cites a company's website but does not recommend that company: the citation does not become a recommendation vote. Similarly: a company can receive a recommendation without its own website being cited. We preserve these distinctions because they measure different kinds of AI visibility.

Mentions Are Also Separate

A company can be:

• Recommended
• Mentioned
• Cited
• Compared
• Criticized

These should not automatically be treated as the same outcome. For standard consensus rankings, the primary event is: meaningful recommendation.

We Use the Same Core Research Question

For a normal consensus study, each platform should receive substantially the same core research prompt. We should not ask ChatGPT:

Which GEO agencies are best for enterprise companies?

and then ask Gemini:

Why is Company A the best GEO agency?

That would invalidate the comparison.

Prompts Should Be Neutral

Standard open recommendation prompts should not seed preferred companies. For example, we prefer:

A B2B SaaS company wants an agency to improve recommendation visibility across leading AI platforms. Which agencies would you recommend and why?

rather than:

Would you recommend CiteWorks Studio, Agency A, Agency B, or Agency C?

unless the study is specifically designed as a named-company comparison.

Related Companies Are Not Seeded Into Standard Prompts

This rule is particularly important for AI Marketing Consensus Index. CiteWorks Studio and LLM Authority Index have related business relationships with this project. They are not automatically inserted into ordinary open-recommendation prompts. If the AI platforms naturally surface them, those recommendations are recorded. If they do not, no recommendation is added.

Competitors Are Not Seeded Either

Neutrality applies in both directions. We also do not construct ordinary prompts to force:

• Profound
• Semrush
• Surfer
• Scrunch
• Otterly
• Writesonic
• Or any other competitor

into the recommendation set. The AI platform determines which companies it surfaces.

Fresh Sessions Where Practical

Prior conversation history can influence AI output. Where practical, standard research should use:

• New sessions
• Minimal prior context
• No unrelated conversation history
• No intentional personalization

This reduces the chance that an earlier mention of a company affects the recommendation.

We Cannot Eliminate Every Form of Personalization

AI platforms may still incorporate factors outside our control. These may include:

• Account settings
• Geographic location
• Platform experimentation
• Personalization systems
• Subscription tier
• Search configuration
• Temporary platform changes

We attempt to control what we reasonably can and disclose what we cannot.

AI Responses Are Not Deterministic

The same platform can provide different answers to the same prompt. A company may appear in one response and not another. Recommendation order may also change. This is an inherent characteristic of generative AI systems.

Consensus Reduces—but Does Not Eliminate—Randomness

Using several AI platforms helps reduce reliance on one individual response. If one company appears independently across six systems, that is a more robust signal than one appearance in one response. However, consensus research does not eliminate model variability.

Research Is a Dated Snapshot

Every study represents the AI recommendation environment at a particular time. For example: September 2026 research should not automatically be assumed to represent: September 2027 recommendations. AI systems can change rapidly.

Platform Behavior Can Change Without Notice

AI companies can change:

• Models
• Retrieval systems
• Search providers
• Ranking behavior
• Safety systems
• User interfaces
• Citation systems
• Training updates
• Product features

These changes may affect recommendation results.

We Preserve Historical Platform Denominators

Suppose an older study used: 7 platforms and a future methodology uses: 8 platforms. We should not retroactively rewrite the old study as though the eighth platform participated. Historical results retain the platform set actually used.

Adding a Platform Does Not Rewrite History

A new AI system may become important enough to add to future research. That does not mean earlier studies were conducted incorrectly. Instead: the methodology evolves prospectively. Historical research remains tied to the platform universe used at the time.

Removing a Platform Does Not Erase Its Old Results

Likewise, if an AI platform:

• Shuts down
• Becomes unavailable
• Stops producing usable responses
• Becomes inappropriate for the research

it may be removed from future cycles. Its historical results should remain part of the record.

What Happens When a Platform Fails?

Sometimes an AI system fails to produce a usable answer. Examples include:

• Technical error
• Timeout
• Refusal
• Incomplete response
• Nonresponsive output
• Answer unrelated to the prompt

That platform may be marked: unusable for that research run.

Unusable Responses Do Not Count as Negative Votes

Suppose seven platforms are scheduled, but only six return usable answers. If five of the six recommend Company A: the recommendation rate is: 5 of 6 = 83.3% not: 5 of 7 = 71.4% A platform failure is not treated as a recommendation against every company.

We Publish the Actual Denominator

Study pages should show information such as: Platforms Scheduled: 7 Usable Responses: 6 That gives readers context for the published recommendation coverage.

We Do Not Claim the Platforms Are Completely Independent

Multiple AI systems can rely on overlapping:

• Websites
• Publishers
• Company documentation
• Review sites
• News articles
• Industry reports
• Public datasets

Therefore: seven AI platforms agreeing does not necessarily mean seven independent source ecosystems agree. It means: seven independently generated platform responses produced similar recommendations.

Shared Sources Are Part of the Research Problem

If several AI systems recommend the same company because all of them repeatedly encounter the same influential source, that is not necessarily noise. It may reveal something important about:

• AI visibility
• Source authority
• Content distribution
• Category positioning
• Citation ecosystems

That is particularly relevant to AI marketing research.

Geographic Differences Can Matter

AI recommendations may differ by:

• Country
• Language
• Region
• Product availability
• Local search results
• Platform availability

Where geography is material to the research, the scenario should state it explicitly.

English-Language Research

Unless otherwise stated, standard AI Marketing Consensus Index research is generally designed around: English-language marketing and business research. Future studies may separately examine geographic or language differences.

Platform Coverage Is Not the Same as Market Share

The seven-platform framework is a research design. It should not be interpreted as a claim that each platform has equal:

• Consumer usage
• Commercial influence
• Traffic
• Revenue
• Market penetration

Equal voting is a methodological choice for measuring consensus.

Why Not Weight by Usage?

Market-share weighting would answer a different question:

What recommendation might the average AI user be most likely to encounter?

Our standard methodology instead asks:

Across the AI platforms studied, how broadly does this recommendation appear?

Both questions can be useful. AI Marketing Consensus Index focuses primarily on the second.

Separate Market-Weighted Research May Be Possible

Future research could potentially examine:

• Platform market share
• Estimated query volume
• User adoption
• Traffic-weighted visibility

If we publish such research, it should be clearly labeled as a separate methodology. It should not be silently mixed into standard Consensus Scores.

Platform-Level Research May Also Be Published

Our structured dataset can support analysis such as:

• Companies ChatGPT recommends most often
• Companies Gemini recommends most often
• Agencies Perplexity recommends most frequently
• Which AI system produces the widest recommendation set
• Where Claude and ChatGPT disagree
• Which platforms cite the most external sources

These analyses are different from standard cross-platform consensus rankings.

Model-Level Research May Be Published Separately

We may also conduct special studies comparing:

• Different models from the same company
• Search vs. non-search modes
• Consumer interface vs. API
• Reasoning vs. standard modes

Those should be labeled: model-level research rather than silently creating additional platform votes.

Related Business Disclosure

AI Marketing Consensus Index has material relationships with: LLM Authority Index and: CiteWorks Studio. LLM Authority Index provides AI research data and measurement infrastructure supporting this project. CiteWorks Studio provides AI search strategy and subject-matter support.

Mark B. Huntley, J.D., who serves as an AI Search & Visibility Research Reviewer, has financial and operational interests associated with these businesses.

These Relationships Do Not Change Platform Results

Related-business involvement does not permit:

• Adding recommendations
• Deleting AI responses
• Increasing recommendation coverage
• Changing recommendation position
• Removing competitors
• Reweighting platforms
• Altering historical data

The raw AI platform responses remain the basis of the recommendation dataset.

Read Our Related Business & Conflict of Interest Disclosure →

LLM Authority Index's Role

LLM Authority Index may provide infrastructure supporting:

• Response collection
• AI recommendation data
• Citation data
• Prompt-level measurement
• Competitive analysis
• Historical AI visibility tracking

When LLM Authority Index itself appears as a candidate in the research: its role as a data provider gives it no additional vote or ranking credit.

CiteWorks Studio's Role

CiteWorks Studio may provide:

• AI search subject-matter input
• GEO/AEO strategy context
• Research taxonomy support
• Research-question design input
• Citation and authority concepts

When CiteWorks Studio itself appears in a study: its strategic contribution gives it no additional recommendation credit or ranking preference.

Our Platform Standard in One Sentence

AI Marketing Consensus Index asks substantially the same neutral question across multiple leading AI platforms, gives each platform one vote, preserves the actual response environment and usable denominator where practical, and measures cross-platform recommendation agreement without allowing market share, commercial relationships, or editorial preference to change the underlying votes.

Frequently Asked Questions

Which AI platforms does AI Marketing Consensus Index analyze?

Our current standard research universe includes:

• ChatGPT
• Gemini
• Claude
• Perplexity
• Grok
• DeepSeek
• Kimi

The exact platform set may evolve over time.

Does every platform count equally?

Yes, for standard consensus research. Each platform receives one vote.

Do you weight ChatGPT more heavily because it is widely used?

No. Standard Consensus Index methodology measures cross-platform agreement rather than usage-weighted exposure.

Do different ChatGPT models count as separate votes?

Not automatically. The standard voting unit is the platform. Model-level comparisons may be conducted separately.

Do you always use the same exact model?

Not necessarily. AI platforms change models and routing systems frequently. Where practical, model and mode information is stored as research metadata.

Do all AI platforms have web access?

Not necessarily. Search and retrieval capabilities can vary by platform, model, mode, interface, and time. Where practical, this is recorded in the research metadata.

Why does web access matter?

A web-connected AI system may discover newer:

• Companies
• Products
• Pricing
• Features
• Reviews
• Research

than a system relying primarily on existing model knowledge.

Are citations counted as recommendations?

No. Citation visibility and recommendation visibility are separate metrics.

Are mentions counted as recommendations?

Not automatically. A company must be meaningfully recommended to receive standard recommendation credit.

What happens if one AI platform fails?

If a response is unusable, it does not count as a negative vote. Recommendation percentages use the actual number of usable platform responses.

Can the same prompt produce a different answer tomorrow?

Yes. Generative AI systems are nondeterministic and can change over time. That is one reason we publish research dates.

Will you add more AI platforms?

Possibly. If another system becomes sufficiently relevant to the research, it may be added prospectively. Historical studies will preserve the platform universe actually used when they were conducted.

Are CiteWorks Studio or LLM Authority Index inserted into the prompts?

Not in ordinary open recommendation research. Related companies must surface naturally unless the study is explicitly designed as a named-company comparison.

Related Pages

How We Rank →

Research Methodology →

Research Definitions & Terminology →

Data & Research Limitations →

Related Business & Conflict of Interest Disclosure →

How LLM Authority Index and CiteWorks Studio Contribute →

Editorial Independence →

About Mark B. Huntley, J.D. →

Transparency

AI Marketing Consensus Index operates with related business relationships that are relevant to our research. AI research data and measurement infrastructure provided by LLM Authority Index. AI search strategy and subject-matter support provided by CiteWorks Studio. Research reviewed by Mark B. Huntley, J.D., where indicated.

These relationships do not alter the underlying AI platform responses, recommendation counts, recommendation coverage, or ranking calculations.

Related business disclosure · Research methodology