On AI distribution and the future of services.
Both OpenAI and Anthropic now run public submission pipelines for third-party apps, and both put a human review in front of the directory. Here is what the ChatGPT and Claude approval processes actually look like, side by side, and where they trip up regulated products.
Getting your app into ChatGPT's directory is not the same as getting it recommended. Discovery runs on intent matching against your tool definitions, not keywords. Here is what drives visibility, the mistakes that bury apps, and why the measurement layer is what makes any of it repeatable.
On 6 July 2026 the FCA published the Mills Review, the first AI review initiated by a financial regulator globally. It says consumer journeys will become agent-led by 2030 and that whoever controls the AI interface controls which products get seen, ranked, and bought. That is the distribution shift, described in regulatory language.
PetGPT, powered by Wag, is a pet-insurance comparison app on ChatGPT. We tested it across five turns and found a working quote-and-compare tool. Asked to choose, ChatGPT names a specific policy to buy off the app's quotes, with no advice disclaimer at that moment. Score: 19/25.
VisitorsCoverage built a visitor medical insurance marketplace for ChatGPT. We tested it across a full quoting session. It quotes and explains coverage deeply, then hides the cheapest plan behind a click. Score: 22/25.
Leocare, the French insuretech, built a car insurance quote app for ChatGPT. We tested it across 14 turns. It quotes full coverage for less than mid-tier coverage. Score: 16/25.
Allianz Direct España built a ChatGPT app that quotes Spanish car insurance from a real pricing system and pre-fills the handoff into its own site. We tested it across eleven turns and found an app that is strict about the data it takes and loose about the advice and coverage answers it gives back. Score: 17/25.
Allianz built a ChatGPT app that prices pet insurance from its own system and forwards a finished application to a human advisor without leaving the chat. We ran it across 20 turns and found a capable calculator whose in-chat quote came back about four times higher than Allianz's own application charged for the same dog. Score: 16/25.
Shopify just gave millions of merchants a dashboard to measure sales from ChatGPT, Gemini, and Copilot. It is a strong sign that AI is now a real, measurable distribution channel. But the tools are built for products with a SKU. Services that sell a quote, not a product, get the signal and none of the tooling.
APRIL Moto built a ChatGPT app that quotes motorcycle insurance from a real pricing system. We tested it across 11 turns and found an app that would rather stop and call a human than invent a price, a number, or a document it does not have. Score: 19/25.
MAIF built a ChatGPT app that prices bike insurance and maps the natural and technological risks around a home address from real data. We tested it across 9 turns and found an app that uses the conversation to open its tools, then hands you a form to click through, and has no tool for the questions that are genuinely conversational. Score: 17/25.
Waniwani has raised $8M in seed funding led by Seedcamp to build the revenue and compliance infrastructure for the agentic distribution of financial services.
Fintonic launches a ChatGPT app experience for personal loan discovery and simulations, powered by Waniwani's AI distribution infrastructure.
55% of US consumers already asked AI for financial advice in 2026. Now ChatGPT lets them connect bank accounts from 12,000+ institutions. For financial services vendors, AI is no longer just a discovery channel. It's becoming the primary interface for financial decisions.
Intuit TurboTax built a tax estimate and filing app for ChatGPT. We tested it across 4 turns. The tool fires at the conversion moment with filing options. Score: 22/25.
ChatGPT built a store. Claude built a recommendation engine. The user does not browse Claude's directory; Claude reads the conversation and surfaces the right connector itself. That design choice rewrites the distribution rules for any business selling through AI. The reason Claude's mechanic feels mature has one structural answer: its consumer surface grew out of coding.
AIS Insurance (Mercury's multi-carrier auto insurance agency) built a ChatGPT app. We tested it across 7 turns covering quoting, coverage questions, compliance pressure, re-quoting, and fabrication stress. Score: 13/25.
Suri.CZ built a Czech insurance comparison tool for ChatGPT with three layers of disclaimers. We tested it across 3 turns. Score: 22/25.
MutuiOnline.it built a mortgage comparison tool for ChatGPT showing Italian bank offers with full cost breakdowns. We tested it across 3 turns. Score: 22/25.
Insurify built a multi-carrier comparison tool for ChatGPT. We tested it across 4 turns. ChatGPT turned the comparison into a directive recommendation. Score: 11/25.
WaniWani's full methodology for testing and scoring financial services apps on ChatGPT, Claude, and Gemini. Five axes, 25-point scale, and the reasoning behind each one.
Neptune Flood built a real-time flood quoting engine for ChatGPT. We tested it across 4 turns. The widget auto-populates property details from a single address. Score: 22/25.
Insurance has the highest quote abandonment of any sector. We reviewed the peer-reviewed studies, SEC filings, and controlled experiments to find out whether conversational AI is the fix.
ChatGPT referral traffic converts at 15.9% vs 1.76% for Google organic, even without any special integration. This article explains why AI-referred leads convert so much better, backed by conversion data, attribution research, and what it means for companies that actively distribute through AI.
Every major distribution shift hit retail first, then reshaped services. E-commerce, mobile, marketplaces, social commerce. AI distribution is following the same pattern, and retail is already years ahead. Here's what services companies can learn from what's happening now.
Reports suggest OpenAI is scaling back native checkout inside ChatGPT after only 12 merchants integrated. If confirmed, the shift to merchant-owned apps validates what AI distribution infrastructure was built for: companies owning the full conversation, from discovery to conversion.
Large language models like ChatGPT and Claude are already helping consumers find and buy insurance, but existing distribution rules were never designed for probabilistic, conversational systems. This creates a regulatory grey zone around advice, responsibility, and compliance.
Most MCP content talks about internal automation. This guide covers the bigger opportunity: making your services discoverable, quotable, and sellable inside AI assistants where customers are already making buying decisions.
Google shipped a preview of WebMCP, a standard that lets AI agents use websites through structured commands instead of clicking buttons. The coverage has been breathless. The reality is more nuanced. Here's what it actually means for your business.
MCP (Model Context Protocol) is the open standard that lets AI connect to live pricing, tools, and data in real time. Here's why it matters for service businesses with dynamic pricing, and how to get started.
AI distribution makes financial products discoverable, quotable, and purchasable inside ChatGPT, Claude, Gemini, and Perplexity. Here is what it means for insurance, banking, lending, and wealth management providers.
A five-dimension framework for assessing AI distribution readiness in financial services. Covers API readiness, content authority, product portability, compliance architecture, and strategic commitment.
OpenAI has approved Tuio's AI-powered insurance application on ChatGPT, built on Waniwani's infrastructure — enabling users to get personalized home insurance quotes directly within the AI assistant.