How to audit a business for AI readiness (2026 guide + free checklist)

Why AI readiness is the next compounding competitive moat

Two things are simultaneously true in mid-2026:

1. AI-mediated buyer research is the new default. Per a16z's consumer AI research, ChatGPT alone now serves hundreds of millions of weekly users for research and purchase comparison. The buyer your business is courting is asking ChatGPT, Claude, Perplexity, and Gemini about you before ever visiting your site. The Anthropic Economic Index on Claude usage shows knowledge-work tasks dominate - including supplier evaluation.
2. Most businesses are faking it. "AI-powered" appears on a third of B2B SaaS landing pages, but the MIT Sloan / BCG AI Maturity research finds the gap between AI-claiming and AI-realising organisations widening, not narrowing. The actual underlying technology is often a single prompt-to-LLM call.

AI readiness audits separate the two. For a customer evaluating you, "uses LangChain or the Claude Agent SDK, has llms.txt, runs Pinecone in production" is meaningfully different from "uses ChatGPT to draft blog posts."

A specific data point worth holding onto: per The Knot's 2025 Real Weddings Study, 36% of engaged couples now use AI to plan their weddings - up from 20% just two years earlier. A 24-month doubling in a famously traditional category. If a category that resists technology that hard is moving 80%+ in two years, the question isn't whether your buyers are AI-mediated - it's how badly you're losing the ones who are.

The 7 signals

1. Customer-facing AI surface

What to check:

• Chatbot platform - Intercom Fin (LLM-powered), Drift GPT, Zendesk Answer Bot, Tidio Lyro, custom widget? Test it: does it answer a hard question intelligently, or just route to FAQs?
• Voice AI - any "speak to our assistant" surface? Platforms: ElevenLabs, Vapi.ai, Bland.ai, Retell.ai, Synthflow, Air.ai.
• AI search / semantic search - site search that understands intent vs keyword? Look for: Algolia AI, Typesense AI, Mendable, Markprompt, Inkeep.
• Personalisation engine - does the homepage / pricing / case studies adapt to the visitor? Look for: Mutiny, Intellimize, Adobe Target, Dynamic Yield, Reflektion.

What most tools miss: the distinction between "has a chatbot" (which Zendesk has had since 2015) and "has an LLM-powered chatbot."

2. AI tech-stack signals - what's actually running

What to check (page source + network panel):

• LLM API calls - OpenAI, Anthropic, Gemini, Cohere, Mistral, Groq, Together AI references
• Agent frameworks - LangChain / LangGraph, LlamaIndex, CrewAI, AutoGen, Haystack, smolagents, Claude Agent SDK, MCP server endpoints
• Vector databases - Pinecone, Weaviate, Qdrant, Chroma, Milvus, pgvector, Supabase Vector, Turbopuffer, LanceDB
• LLM observability - LangSmith, Langfuse, Helicone, Braintrust, W&B Weave, Arize, Phoenix, Portkey

What most tools miss: the observability signal. A business running LLM observability is measuring their AI quality. A business without it is shipping AI features blind.

3. Automation + integration depth

• Workflow platforms - Zapier, Make.com, n8n, HubSpot Workflows, Pipedream, Workato, Tray.io
• CRM signals - Salesforce, HubSpot, Pipedrive, Close, Attio. For solo / creative service businesses, HoneyBook is the equivalent.
• Scheduling - Calendly, SavvyCal, Cal.com, Acuity Scheduling
• Lead routing - HubSpot, Pipedrive, Typeform routing. A "Contact us → 24-48 hour response" flow without automation is a real signal of low AI maturity.

4. Data + analytics maturity

• Analytics - GA4, Mixpanel, Amplitude, Heap, PostHog. Multiple tools = mature; single-tool dependence = weaker.
• CDP (Customer Data Platform) - Segment, RudderStack, Hightouch, Census. CDP presence is the single best signal of data maturity.
• Data warehouse - BigQuery, Snowflake, Databricks, Redshift mentioned in job postings, engineering blog, or DPA
• Event tracking depth - does the JS load custom events (e.g. analytics.track('signup')) or just pageviews?

What most tools miss: CDP detection.

5. Emerging signals (2024-2026)

• llms.txt - the Answer.AI standard, adopted by Anthropic, Cursor, Perplexity. Live at /llms.txt.
• Agent-shoppable schema - complete Product + Offer + GTIN + Brand + availability JSON-LD for ChatGPT Shopping, Perplexity Shopping, Google Shopping AI.
• AI governance / disclosure page - EU AI Act Article 50 (effective Aug 2025) requires disclosure of AI use to consumers.
• AI training opt-out - distinct from cookie consent. GDPR Article 22 creates a separate consent surface.

6. The claim-vs-reality gap

What to check (compare marketing copy to evidence):

• Page says "AI-powered chatbot" → is there an LLM API call in the network? Or is it Zendesk's rule-based bot from 2018?
• Page says "agentic AI workflows" → are LangGraph, CrewAI, or MCP server signals present? Or is it a Zapier chain renamed?
• Page says "personalised at scale" → is there a real personalisation engine? Or is the same homepage shown to everyone?
• Page says "AI-native" → does the team have AI/ML engineers in job postings, or just a marketing manager who learnt ChatGPT?

7. Competitor AI adoption

If your closest competitor has a working LLM-powered chatbot, llms.txt, agent-shoppable schema, and you don't - you're losing AI-mediated buyer evaluations. Pick 3 direct competitors. For each, score the 6 signals above. Compare against your own score. Same comparison logic applies on the conversion side - see the marketing conversion audit guide for the funnel-side comparison framework.

Worked example - a real SMB we audited

Composite score: 28/100 - Level: AI Unaware

The business is a small service operator. Modern website, GA4 + Consent Mode v2 correctly implemented, proper Schema.org LocalBusiness markup, active social media. The technical foundations are above average for a solo operator. As covered in the privacy audit guide, the same business's privacy posture is also a C - strong foundations, fixable gaps.

But the audit found:

• Zero customer-facing AI. No chatbot. No live chat. No semantic search. No personalisation. A bride landing on the site at 10pm Friday gets nothing until Monday morning. The site's own contact page admits a "24-48 hour" response window.
• Zero AI tech-stack signals. No LLM API calls in the source. No agent framework. No vector DB. No observability. No MCP endpoints.
• Automation depth: none visible. Every step from enquiry to booking is manual. No Acuity, Cal.com, Calendly, HoneyBook, Mailchimp, Zapier, or n8n.
• Data maturity: foundations only. GA4 with consent mode - positive. But no CDP, no event tracking beyond pageviews, no email list, no CRM. 6 Google reviews total - too small a sample to feed any model.
• Emerging signals: all absent. No llms.txt. No agent-shoppable schema. No AI disclosure page.
• Claim-vs-reality gap: none. Genuinely zero AI claims in the copy. Honest absence - which is better than dishonest presence.
• Competitor adoption: the three named local competitors are at the same maturity level. None have chatbots, online booking, or AI features. The category is structurally behind the industry trend.

Score breakdown

30 / 60 / 90 day roadmap (tool prices as of May 2026)

At a conservative $50/hour, reclaiming 15-26 hours/month against ~$117/month in tool costs is $750-$1,300 of effective time savings, every month. And in a category where 36% of buyers are AI-mediated, being the first local artist with a working chatbot is genuinely a first-mover advantage. The same business's reputation gap (covered in the reputation audit guide) compounds this - AI engines preferentially cite businesses with strong review signals and structured data, so AI readiness + reputation work together.

Run it on your site automatically

FAQ

Written by Adam Savage, Founder of AuditHQ - 17 years building digital products. Last updated 13 May 2026.