For Marketplaces & Enterprise

We measure how AI engines recommend your category. Then we engineer the citation share that follows.

Across ChatGPT, Claude, Gemini, and Perplexity. Per-query, per-engine, per-landing-URL, per-intent-class breakdown of AI engine retrieval behavior - joined to your first-party server logs, your paid CPL data, and your CDP.

We extend your analytics stack. We don't displace it. No new platform required. Your team gets the dashboards and the quarterly refresh.

What we don't measure: training-data exposure. We measure citation behavior - what AI shows users when they answer category queries. Anyone claiming to measure training-data exposure directly is overpromising. The methodology and what we explicitly don't claim are named in writing in every engagement.
Audit substrate

Standard scope. Configurable for vertical depth.

65
Buyer queriesMapped to your intent classes - Direct Brand, Comparison, Discovery, Local, Purchase Intent, Competitive Displacement.
4
AI enginesChatGPT, Claude, Gemini, Perplexity. Cross-engine consensus is the signal.
260
AI responses capturedPlus retrieval-side architecture probes and AI bot-behavior modeling against published bot IP ranges.
3
Reliability tiersHIGH (directly observed), MED (statistically inferred), LOW (industry intel). Every claim tagged.
Engines covered ChatGPT Claude Gemini Perplexity

Built for

Enterprise operators who skim, judge methodology, and don't tolerate vendor displacement.

28 Labs Enterprise is scoped for organizations with multi-hundred-million-dollar P&Ls, internal engineering and data science teams, mature analytics stacks, and partnership-exposed revenue worth defending. The audit substrate, the integration model, and the deliverable shape are tuned for that buyer.

Marketplaces

Auto, real estate, jobs, travel, financial services - any multi-sided marketplace where AI engines now compose category answers (lender lists, vehicle valuations, neighborhood comparisons, partner shortlists) that route demand around your partnership corridors.

Enterprise B2C & B2B

Companies where AI is becoming a measurable referral source, Adobe RT-CDP or equivalent is already in production, and paid CPL is being scrutinised by finance teams that need a reallocation case backed by query-level evidence.

Partnership-exposed revenue

Where exclusive financing, distribution, or fulfilment partnerships represent a high-value commercial asset, and the offline corridor doesn't reproduce on AI surfaces - because AI engines compose competitor lists side-by-side rather than honour the exclusivity.

CFO-grade orgs

Where every recommendation needs to land at CPO / CTO / CDO level, be forwardable to the board, and survive scrutiny from internal data science teams. We brief at that bar.

The method

Four layers, one substrate.

Each layer extends the one below it. The audit substrate is delivered in every engagement. The three intelligence layers above it are scoped to client need - and operate against existing client infrastructure.

L1 - SUBSTRATE

AI Discovery Audit

The measurement substrate. Standard scope 65 buyer queries x 4 engines = 260 AI responses captured. Architecture probes and bot-behavior modeling layered on top.

  • Per-query, per-engine, per-landing-URL, per-intent-class behavior
  • Retrieval-side probes: robots.txt, rendering posture, CDN geo-policy
  • AI bot-behavior modeling against published bot IP ranges (GPTBot, ClaudeBot, PerplexityBot, ChatGPT-User, Claude-User)
  • Five-dimension AI Recommendation Score: Top Recommendation Rate, Visibility Rate, Funnel Coverage, Site Readiness, External Presence
L2 - INTELLIGENCE

Citation Intelligence

Per-URL citation share across every retrieved AI engine. Where competitors are cited above you. Where the answer-engine format pattern is already working inside your own portfolio - and where it isn't.

  • Per-URL citation counts across all 4 engines
  • Format-pattern analysis: which content shapes AI engines retrieve from, which they skip
  • Competitor citation share within your category
  • Authoritative-source ratio: your own pages vs the institutional source on your own data
L3 - OBSERVABILITY

First-Party Log Intelligence

We integrate your server-log feed, classify AI bot traffic, and surface live AI-channel demand signal. This is substrate Adobe Analytics, Target, and Audience Manager structurally cannot produce - the bots don't execute the JavaScript those tools depend on.

  • Per-bot crawl behavior (verified against published IP ranges - user-agent spoofs filtered)
  • Live AI prompt signals: ChatGPT-User, Claude-User, Perplexity-User fetches = real human demand
  • AI referral attribution joined to Adobe Analytics (per-engine landing distribution + funnel)
  • Quarterly refresh; your team gets the dashboards
L4 - COMMERCIAL

AI-Paid Substitution Matrix

We ingest your paid spend feeds and CDP conversion data, classify the query universe at the intent-class level against the AI demand we measure in the audit substrate, and produce a per-category reallocation matrix in dollar terms.

  • Four-quadrant framework: AI demand x paid CPL efficiency x CDP conversion
  • Reallocation candidates with dollar-quantified opportunity per category
  • AI traffic treated as incremental, not as one-for-one paid replacement
  • Reallocation fires only on the bleeding subset where CPL is at or above target
  • Refresh quarterly. Your team or paid agency executes; we operate the measurement.

Methodology discipline

Every load-bearing claim carries a reliability tier.

No mystery numbers, no hand-waving. Internal data science teams should be able to audit any cited figure back to its method. The reliability framework is shown on slide 1 of every executive review.

HIGH

Directly observed

Captured from the live AI engine response or measured against a verifiable artifact: a curl probe, a published bot IP range, a robots.txt directive, a citation in the actual AI output. Reproducible.

MED

Statistically inferred

Derived via a documented statistical method (intent attribution from log patterns, format-pattern clustering, cross-engine consensus modeling). Methodology disclosed; honest about the inference layer.

LOW

Industry intel pending data

Estimated against published industry signal, not yet joined to your specific data. Marked LOW explicitly so the joined substitution-matrix output (with your CPL + CDP feed) is the trusted version, not the placeholder.

Integration posture

We plug into your stack. We don't displace it.

No new analytics platform. No vendor lock-in. No "you need our SaaS" pressure. The intelligence layers operate as a managed service against existing client infrastructure.

Server-log feed

Existing access logs (CloudFront / nginx / Akamai / Fastly). We classify AI bot traffic against published IP ranges and turn raw logs into three actionable intelligence streams. No new instrumentation.

Adobe RT-CDP (or equivalent)

The Substitution Matrix joins to your existing CDP. We don't replace it. We classify the query universe at the intent-class level and produce reallocation candidates against your real conversion data.

Paid spend feeds

Whatever paid platform your team or agency runs (Google, Meta, programmatic). We ingest the spend feed and the CPL data, the agency continues executing. We operate the measurement layer.

Adobe Analytics / Target / Audience Manager

AI referral attribution lands in your Adobe Analytics so per-engine landing distribution + downstream funnel behavior become observable inside the tool your team already uses every day.

Internal engineering teams

28 Labs specifies the page architecture and re-measures citation share against the baseline. Your engineering team builds. We don't deploy to your codebase.

Why this matters

AI bots don't execute the JavaScript that Adobe Analytics, Target, and Audience Manager depend on. The behavior never reaches those tools. The substrate sits in the server logs, not the analytics layer.

28 Labs delivers the AI-channel layer as a managed service feeding intelligence back into the platforms your teams already operate. You get the answer-engine signal, in the tool you already trust, on a quarterly refresh cadence.

Your CISO will appreciate that we deliver intelligence as an outbound feed and dashboards, not as an embedded SDK. No new attack surface, no new vendor inside the production rendering path.

See what a real audit looks like

An anonymized executive summary from a recent enterprise marketplace engagement. 65 queries x 4 engines = 260 AI responses, joined to retrieval-side architecture probes. The same shape every audit produces - methodology, scorecard, the structural-leak finding, the substitution-matrix shape, the pilot scoping.

Open the Sample Audit

Engagement model

Three ways to engage. Sized to where you are.

Indicative scope below. Actual engagement structure is shaped in the working session against your category, your engines, and the analytics integration that already exists.

Next step

A 60-minute working session.

We walk a sample audit, the substitution-matrix shape for your category, and the strategic options for which class of move is most actionable first. No deck-and-pitch. Working session. We come in with a category read.

28 Labs is operated out of Dubai (Lab 28 Technologies - FZCO). Engagements are signed on DocuSign. Methodology and what we explicitly do not claim are named in writing.