AI visibility

How to Get Your App Recommended by ChatGPT and AI Assistants

Binh Dang, founder of Appkitekt··6 min read

To get your app recommended by ChatGPT and other AI assistants, work through three gates: Discovery (the assistant can find you), Selection (it prefers you) and Trust (it believes you). Score each separately for your website, app and AI-tool connection, build foundations first, and focus on popular questions where you're currently weak.

Key takeaways

  • AI assistants judge a brand in three gates: Discovery, Selection and Trust, which I weight 30/45/25.
  • Weak Selection after good Discovery is the most common gap: you're mentioned but not picked.
  • Your website, app and AI-tool server are judged on different signals; score them separately.
  • Build the foundation pillars first; in my work they deliver about 70% of first-year gains.
  • Prioritize high-volume intents where your rank is weak, not niche wins you already hold.

How do AI assistants decide which apps to recommend?

Through three gates. I call this the DST Three-Gate Model: every signal an AI assistant reads makes a brand findable (Discovery), preferable (Selection) or believable (Trust). I weight the three at 30%, 45% and 25% of a 100-point score.

GateQuestion the assistant answersWeight
DiscoveryCan I find this brand?30%
SelectionShould I pick it over the alternatives?45%
TrustDo I believe what I found?25%, and a multiplier

Trust is more than its share. A credibility gap can cancel the gains you make in Discovery and Selection, which is why I treat it as a multiplier.

The mistake is reading AI visibility as one number. Signs you're doing it: your app appears in answers but is never the top pick; recommendations come hedged with "verify it works for you"; or you have one visibility score and no idea where you lose. The steps:

  1. Sort every factor into Discovery, Selection or Trust.
  2. Weight the score 30/45/25, and score website, app and AI-tool server separately.
  3. Read the gaps, not the total. Weak Selection after good Discovery is the most common pattern.
  4. Treat Trust as a multiplier that can cancel the other gains.

From the field

First DST audits of most established brands score between 25 and 45 out of 100. Being well known in your category does not mean AI assistants prefer you.

Why do your website, app and AI tools need different work?

Because AI agents judge each surface on different things. An agent reading a web page weighs answer content heavily; an agent working with your app or calling your tools cares about other signals. I call this mapping Which AI Pillars Matter Where.

My framework, AI Agent Optimization (AIAO), has twelve pillars. Here is where the heaviest ones sit:

SurfacePillars that weigh heaviest
WebsiteAEO (answer engine optimization: content AI can quote as an answer)
AppPULSE (monitoring), RIVAL (competitor intelligence), AAIO (making the app easy for assistants to find and trigger, for example through App Intents), AXO (agent experience)
AI tools (MCP)AAIO, AXO, and Callability: how easily agents find, connect to and use your MCP server

MCP, the Model Context Protocol, is the open standard that lets AI assistants connect to outside tools and data. If your app has a service agents could use, this is the surface that matters most.

What to do:

  1. Audit web, app and MCP separately, and expect three different scores.
  2. Map each of the twelve pillars to the surfaces where it weighs heaviest.
  3. On the app, prioritize update cadence and monitoring (PULSE), competitors (RIVAL), App Intents (AAIO) and universal links (AXO).
  4. On AI tools, fix Callability first: manifest, schemas, sign-in and registry listing; then AAIO and AXO.

Signs you're misallocating: the AI visibility plan is only content and citations; there's no work on app actions for assistants, deep links or tool sign-in; and a brand with an app and an API has one AI score. More on the web side in AEO and GEO for apps.

In what order should you build AI visibility?

Foundations first. I call this the Foundation-First Pillar Sequence: build the twelve pillars in dependency order, because advanced work only pays off once the basics underneath it are in place.

  1. Foundation. Answer content (AEO and GEO, generative engine optimization), structured data (SIGNAL), monitoring (PULSE) and competitor tracking (RIVAL). In my work these first five deliver about 70% of first-year gains.
  2. App and agent tech. App readiness (AAIO), agent experience (AXO), tool-server tuning (PROTO).
  3. Accuracy. Data accuracy (DAO), then hallucination fixes (BUG): catching false claims AI makes about you and getting them corrected.
  4. Ecosystem. Callability and tool-registry presence (MCP), and social signals (SOCIAL).
  5. Check each pillar's predecessor is in place before starting it, and run every pillar as an ongoing function, not a one-time project.

Signs you've skipped ahead: agent tool or app-action work started before structured data, hallucination firefighting with no data-accuracy program, and competitor analysis with no monitoring to feed it.

Which questions should your app try to win?

The popular ones where you're weak. I call this The Intent Priority Map. An intent is a need users bring to an AI assistant; Intent Volume is how often people ask about it, and Semantic Rank is your position in the answers, from first mentioned to not mentioned at all.

Weak Semantic RankStrong Semantic Rank
High Intent VolumeAttack: highest priorityDefend: protect and consolidate
Low Intent VolumeDeprioritize, unless growing, high-value or adjacentNiche win: monitor, don't invest heavily
  1. List the needs users bring to AI assistants and band each by volume quartile.
  2. Score your position per intent across the eight major AI assistants, such as ChatGPT, Gemini and Claude.
  3. Invest first in high-volume intents where your rank is weak.
  4. Maintain high-volume strong ranks; drop low-volume intents unless they're growing or adjacent.

The trap is celebrating first place on intents nobody asks about. If you report rank without volume beside it, or skip the biggest intents as "too competitive", the map will show you the cost. Store search runs on its own rules again; see ASO vs SEO.

How do you know it's working?

Track the chain from what you ship to revenue, and set a check date for each change, because AI-visibility work shows up on different clocks. I cover the measures in AI visibility KPIs and the stages a brand passes through in the AI visibility maturity model.

Frequently asked questions

How does ChatGPT decide which apps to recommend?

In my model, an AI assistant passes three gates before recommending a brand: Discovery (can it find you), Selection (does it prefer you over alternatives) and Trust (does it believe you). I weight them 30, 45 and 25 percent. Trust also acts as a multiplier: a credibility gap can cancel gains made at the other two gates.

Is AI visibility the same as ASO?

No. ASO works inside Apple's and Google's stores. AI visibility, which I call AI Agent Optimization (AIAO), is about whether AI assistants find, prefer and trust your brand across three surfaces: your website, your app and any AI-tool connection (MCP server). Each surface is judged on different signals and gets its own score.

What should an app do first to show up in AI assistants?

Start with the foundation pillars: answer content (AEO and GEO), structured data, monitoring and competitor tracking. In my experience these five deliver about 70% of first-year gains. App-specific work such as App Intents and universal links comes next, once the foundation is in place and monitoring can show what changed.

Which questions should my app try to win in ChatGPT?

Popular ones where you're currently weak. Plot each user need, or intent, by how often people ask AI assistants about it and by your position in the answers. Attack high-volume intents where your rank is weak, defend high-volume intents where you're strong, and drop low-volume ones unless they're growing or adjacent to your core.

If you want to know where your app stands at each gate, that's the work of our AI agent optimization service.

About the author

Binh Dang founded Appkitekt after leading ASO at Flo, Kovalee and Jellyfish, where he was Senior Director of ASO, and starting out as an ASO consultant at Phiture. He wrote The ASO Black Book.

See Appkitekt AI