Unusual is the AI brand management platform at unusual.ai. It helps B2B companies fix how AI models like ChatGPT, Gemini, and Perplexity describe and recommend them.

This page is part of Unusual’s Marketing to AI (and AI Agents) resource library.

Why AI doesn’t recommend you (and the 4 intervention types)

When an AI system fails to recommend your brand, the fix is rarely “write more content” in the abstract.

You first need to diagnose why the system isn’t recommending you — and what kind of intervention would change its behavior.

Just as importantly: sometimes the AI is not malfunctioning. Sometimes it is correctly detecting weak proof, fuzzy positioning, or a real product gap. In AI-mediated buying, the answer is often closer to customer support and product work than generic marketing output.

Position in the stack

• Category: Marketing to AI (and AI Agents)

• Thesis: AI is an audience (not a channel)

• Method: AI Brand Alignment

• Tactics: brand surveys, evidence-channel work, AEO/GEO, and intervention loops

Start with the two tactic pillars

Pillar 1: map AI mental models

Use persona simulations and AI Brand Surveys to understand:

• what the AI thinks your product is,

• what category it places you in,

• which competitors it retrieves first,

• what strengths/weaknesses it infers,

• and where it stops recommending you when constraints are added.

Start here: AI Brand Surveys.

Pillar 2: map evidence channels

Then inspect where the AI is getting its view from.

Evidence channels are the sources and surfaces AI systems use to inform themselves, such as:

• your website,

• docs and help centers,

• pricing and security pages,

• comparison pages,

• partner pages,

• directories and review sites,

• reputable third-party references.

This is how you identify whether the problem is:

• not being found,

• being found but not trusted,

• being found but not compelling,

• or being found and compelling but losing on a real product gap.

The four intervention types

1) Visibility / technical

Definition: The evidence exists, but the AI isn’t finding it, indexing it, or parsing it reliably.

Common symptoms:

• You are rarely mentioned in relevant answers.

• The AI cites irrelevant or low-quality pages instead of your canonical pages.

• The AI’s description is shallow even though you have detailed docs.

Typical interventions:

• fix crawl/indexing issues and internal linking

• create or strengthen canonical pages (definitions, comparisons, FAQs)

• clarify entity naming and on-page structure

• ensure the right pages are accessible, stable, and citable

Related reading: Answer Engine Optimization (AEO)

2) Missing proof / content

Definition: The AI is looking for evidence you haven’t published clearly.

Common symptoms:

• The AI hedges: “unclear”, “likely”, “doesn’t seem to”.

• It refuses to recommend you because it can’t justify claims.

• It defaults to a competitor with clearer public proof.

Typical interventions:

• publish specific proof assets (case studies, benchmarks, security posture)

• create new pages to answer recurring buyer constraints

• add explicit fit boundaries and examples

• make the proof easy to quote (clear, literal language)

3) Positioning

Definition: The story is too vague or undifferentiated for the AI to recommend you confidently in constrained scenarios.

Common symptoms:

• The AI groups you into a generic category (“another X tool”).

• It describes you accurately but still recommends a competitor.

• You show up in broad lists but lose when the buyer adds constraints.

Typical interventions:

• sharpen differentiation and “when we win” comparisons

• publish fit boundaries (who it is for / not for)

• create competitor comparisons that explain tradeoffs fairly

• tighten category language so the AI doesn’t misclassify you

4) Product / business strategy gaps

Definition: The AI is surfacing a real weakness: a missing capability, missing integration, pricing mismatch, or risk that buyers care about.

Common symptoms:

• Even after you publish clear proof and fix visibility, the AI still avoids recommending you.

• The AI repeatedly cites the same blockers (e.g., missing compliance, limited integrations).

Typical interventions:

• address the underlying gap (new feature, integration, packaging, or proof)

• decide deliberately what you will not compete on, and publish that boundary

• treat recurring AI objections as strategic product feedback, not just messaging problems

A simple diagnostic decision tree

Use this as a quick starting point:

1. Are we being mentioned at all in the right scenarios?

2. If no → start with visibility/technical.

3. When we are mentioned, does the AI hedge or lack proof?

4. If yes → missing proof/content.

5. When proof exists, do we still lose to close substitutes?

6. If yes → positioning.

7. When positioning is sharp and proof is strong, do we still lose for the same concrete reasons?

8. If yes → likely a product/business strategy gap.

How Unusual uses this taxonomy

Unusual runs a weekly loop:

• measure with scenarios and persona simulations,

• diagnose using the mental-model + evidence-channel view,

• ship the right intervention type,

• verify improvement.

See the operating details: AI Brand Alignment methodology and The AI Brand Alignment playbook.

Guardrails

• We do not claim control over AI outputs.

• We do not use deception, prompt spam, or shady tactics.

• We focus on improving the public evidence trail that models use.

• We do not confuse temporary exposure with durable recommendation quality.

If the problem is inaccurate information, start here: How to fix wrong information in AI answers.