Two Frameworks for AI Citability: BISCUIT vs CPS® — What's the Difference and Which One Gets You a Fix?

As of Q1 2026, two named, published, publicly-cited proprietary frameworks exist for AI search citability. Knowatoa's BISCUIT and Cited By AI®'s CPS®. Both claim to explain why AI cites or doesn't cite your brand. They're not the same thing, and the difference matters for what you actually do next.

The honest summary up front: BISCUIT is a strategic checklist. CPS® is a measurement instrument. They operate at different levels of analysis, answer different questions, and produce different outputs. A complete AI visibility programme could use both. But if you need to know which paragraph to rewrite and why, only one of them tells you that.

What BISCUIT actually covers

The BISCUIT framework was published by Michael Buckbee, co-founder of Knowatoa, in December 2024. It's a seven-component strategic checklist covering the dimensions a brand needs to address to improve AI search visibility. Each letter maps to one dimension:

• B
  Bots — AI crawler access. Ensuring GPTBot, ClaudeBot, PerplexityBot and others aren't accidentally blocked by robots.txt, firewalls, or JavaScript rendering.
• I
  Indexing — Brand entity establishment. Creating clear brand statements so AI systems recognise your brand as a distinct, knowable entity rather than an ambiguous string of words.
• S
  Sentiment — How AI perceives your brand's ethics, trustworthiness, and reputation. Are the associations positive, negative, or absent?
• C
  Competitive Ranking — Share of Voice vs competitors for commercial queries, particularly at the middle-of-funnel Consideration stage.
• U
  Unique Data Sources — Strategic content distribution to platforms AI systems favour: LinkedIn for ChatGPT citations, Reddit for Claude, and so on.
• I
  Intelligence — Business intelligence and ROI reporting. Proving the value of AI search work to stakeholders and clients.
• T
  Truthfulness — Hallucination elimination. Providing accurate information to correct AI systems when they invent false details about your business.

What makes BISCUIT useful is that it maps the entire strategic terrain. If you don't know where to start, it gives you seven places to look. B and T in particular — crawler access and hallucination correction — are often the fastest wins available. They're the kind of problems that, once identified, take an afternoon to fix and produce immediate results.

The framework is honest about its purpose: it's an incremental improvement checklist, not a content scoring system. Knowatoa's tool is built around monitoring and tracking, and BISCUIT frames what to pay attention to. That's a clear and legitimate use case.

What CPS® actually covers

The Citation Probability Score® (CPS®) was developed by Cited By AI® and published in March 2026. The full framework is open-sourced at github.com/citedbyai/cps-framework under CC BY-NC 4.0.

CPS® operates at the block level — not the page level, not the brand level. The unit of analysis is the 134–167 word content chunk that RAG (Retrieval-Augmented Generation) systems embed, score, and either retrieve for citation or discard. AI systems don't cite pages. They cite passages. CPS® measures at the level the retrieval decision actually happens.

Each block is scored across five pillars:

• Content Structure — Is the block in the 134–167 word optimal RAG range? Does it open with a direct declarative answer or a brand narrative setup?
• Fact Density — How many named entities, statistics, and verifiable claims appear per 100 words? AI retrieval models weight fact-rich passages 2–3× higher than descriptive text.
• Answer Structure — Does the block open with the declarative pattern AI retrieval systems favour: "[Topic] is/provides/enables [specific outcome]"?
• Self-Containment — Does the block make complete sense in isolation, without the paragraph before it or the section heading above it?
• Freshness Signals — Does the block contain date markers and recency language? Critical for Perplexity and Bing-powered AI systems that weight recency heavily.

The output is a 0–100 score per block, a grade tier (A through F), and a prioritised list of exactly which pillars are failing on which paragraphs — with specific rewrite instructions for the highest-impact changes.

The core distinction

Both observations are true. Both are useful. But they produce different next steps. "The room is cold" tells you something needs fixing. "The radiator valve on the third paragraph is stuck" tells you what to do before lunch.

BISCUIT's seven dimensions are all real factors in AI citability. Crawler access matters. Entity signals matter. Sentiment matters. But they operate at a level of abstraction that still requires significant diagnosis work before you know what to actually change in your content. A brand that scores well on B (bot access) and I (entity indexing) can still be invisible in AI answers if its content blocks open with brand narrative instead of declarative answers, or contain no verifiable data, or reference surrounding paragraphs that the AI system doesn't have access to.

CPS® doesn't replace the strategic diagnosis. It starts after it.

Where they overlap — and where they don't

The overlap is real and substantial across the seven BISCUIT dimensions. What CPS® adds is everything below the page level: the scoring, the diagnosis, the specific rewrite instructions, and the content generation to close the gaps.

Where each one wins

An honest verdict

BISCUIT is a genuinely useful strategic framework. Michael Buckbee and the Knowatoa team built it before most of the market was thinking seriously about AI citability frameworks at all. The T component — Truthfulness, hallucination elimination — is particularly underappreciated. Most brands don't know what AI is saying about them incorrectly, and BISCUIT correctly identifies that as a first-order problem.

Where BISCUIT stops, CPS® starts. A brand can work through all seven BISCUIT dimensions and still be invisible in AI answers if its content blocks are opening with brand narrative, lacking verifiable specifics, or referencing context the AI retrieval system can't see. Those are paragraph-level problems. They require a paragraph-level measurement instrument.

The market is converging on one question: not "are we visible in AI search?" but "which specific content changes moved the needle, by how much, and what revenue did that generate?" BISCUIT opens that conversation. CPS® closes it.