FAQ

AI search and conversion optimization — but full-funnel (Visibility → Recommendation → Referral → Handoff) and full-stack (audit, optimize, measure, monitor, attribute, host MCP, coach, consult) instead of picking one point on the funnel or one tool in the stack.

Getting cited by ChatGPT is stage 1 of 4. Appkitekt AI tracks whether the agent then recommends you (stage 2), whether that recommendation drives a referral (stage 3), and whether the referral converts into a signup, purchase or tool call (stage 4). Optimizing one stage in isolation just moves the leak downstream.

Because a monitoring-only tool tells you you're losing without fixing it, a hosting-only tool ships tools nobody measures, and a consulting-only vendor bills you to write playbooks that go stale in a month. AIAO ships audit, optimization, measurement, monitoring, attribution, hosted MCP, coaching and consulting on one wallet — so the loop closes weekly instead of quarterly.

Within a category, yes — the rubrics are the same. Across categories, treat it as directional. A SaaS with an API and a consumer content site are optimized against different weights.

Each model weighs signals differently. Schema density, freshness, structured endpoints — a fix that lifts you on one model may barely move another. That's why Optimizer recipes are agent-specific.

Push-to-Live deploys in under 60 seconds on web. The agents themselves re-crawl on their own schedules — typically 24 hours to a week. The next AIAO audit cycle measures the lift.

No. We guarantee the fix ships correctly and is technically valid. Agent behavior is non-deterministic, and we don't claim to control it.

Yes. Every audit and intelligence report exports to CSV and JSON. Higher tiers expose a webhook.

The button greys out and tells you which connection is missing. Wire it from the Connections panel and the pillar unlocks.

We store the prompts (they're authored by us) and the responses (so trends are reproducible). We do not store any of your end-user data — AIAO never sees it.

A self-built setup tells you what one model thinks. AIAO tests six, runs them on a fixed cadence, and ships agent-specific fixes. The comparison page in /why/comparisons covers this in detail.

Paste an App Store or Play listing link. We resolve it to a bundle id and listing data, then run the same DST chain on app-specific signals — listing structure, what's-new, AASA, App Intents, ratings velocity, rival-app overrides.

Read-only audit works from public listings. Push-to-Live store-listing drafts require an API key with the right scopes. Without it you stay in read-only mode and we surface fixes as repo PRs and JSON files you can ship manually.

Push-to-Live stages metadata drafts in App Store Connect / Play Console and repo PRs for AASA, assetlinks and App Intents. You approve, Apple or Google index on their normal cycle (typically 24h–7d), and the next audit measures the lift.

Different surface, different factor floor. Both roll up to the same 0–100 scale — the floor only affects the number we cite in copy.

It complements it. Classic ASO tools optimize for store search; AIAO optimizes for agents recommending the app outside the store. Run them together.

Yes. Paste the manifest URL (or the domain hosting /.well-known/mcp.json) and Appkitekt AI runs a 210+ factor audit covering registry presence, manifest validity, tool schemas, auth scopes, error shape, latency and per-agent hallucinated-call rates. Same 0–100 Index, same agent panel — only the prompt class flips to tool-calling.

Tool DST score replaces the headline Index label, Tool-call share replaces Mindshare, plus Registry coverage, Tool-selection precision, Schema fidelity, Latency p50/p95 per tool, Token footprint and Hallucinated-call rate. Reading the numbers covers the full list.

Manifest fixes ship as repo PRs against /.well-known/mcp.json, OpenAPI specs and tool description files. Registry submissions are staged for one-click submit to Smithery, MCP-Index and the Anthropic registry. Agents pick up the new manifest on their normal refresh cadence and the next audit measures the lift.

On web and app, MCP is one channel among many — agents have HTML, schema and listings to fall back on. For pure tool-server vendors the MCP server is the surface, so the MCP pillar weights up to ~16% and AAIO/AXO/PROTO weight up alongside it. The full per-mode weights are in Scoring & weighting.

No. Both draw from the same credit wallet as concierge chat, audits, mindshare, and Optimizer. Your monthly retainer plan tops up the wallet; the wallet powers everything.

Ingest and dispatch pause — Attribution collector and MCP tool calls return HTTP 429 with an 'add credits' hint. Nothing is silently overcharged. Read cached reports and dashboards are always free, so the app stays usable while you top up.

Attribution ≈ 0.0002 credits per event (100k events ≈ 20 credits). MCP ≈ 0.02 credits per tools/call (1,000 calls ≈ 20 credits). Storage past 30 days is a tiny fractional credit per unit per day. Full rates + worked examples: /guides/credits.

Yes. The retainer is not a feature gate — every capability is unlocked on every plan. What the retainer gives you is a predictable monthly credit allowance at a better per-credit rate.

Human-in-the-loop guardrail. Any autonomous change — MCP spec-drift remediation, Attribution auto-heal, or Optimizer push-live — is queued as a change request. A team owner approves or rejects before it ships. Rejected requests do not consume credits.