# When customers ask AI, are you in the answer? > Winin reveals how AI discovers, understands and recommends your brands and products—then turns each gap into evidence-led work your team can approve, publish and retest. - Canonical: https://winin.ai/ - Updated: 2026-09-04 Winin · AI RECOMMENDATION INTELLIGENCE Your customers are already asking AI. When customers ask AI, are you in the answer? Winin reveals how AI discovers, understands and recommends your brands and products—then turns each gap into evidence-led work your team can approve, publish and retest. Illustrative interface · sampled by agreed questions, models and cadence The monitored model roster includes ChatGPT, Claude, Gemini, Perplexity, DeepSeek, Grok, Copilot, Kimi, Qwen, GLM, Doubao, MiniMax and ERNIE Bot. 01 · The Problem ## Before customers make a purchase decision, AI may already be recommending a competitor. These conversations happen every day — you simply aren’t there. CUSTOMER ASKS One question · four possible outcomes ↓ 01 ABSENT Each absence can reduce the chance that customers include your brand in their consideration set. 02 WRONG Stale or incorrect information may reach an answer before your correction and appear under your brand name. 03 NOT RECOMMENDED AI: “It’s also decent. But as for reasons —” every one listed belongs to competitors. Being mentioned is not being chosen. What customers remember are someone else’s reasons. 04 BLIND SPOT Without visibility there is no response; budget spent, with no way to verify the effect. 02 · The Gates ## An AI recommendation is not a switch. Across three stages, a brand can be discovered, enter consideration and receive an accurate recommendation. Winin identifies where the gap occurs and prioritizes work supported by evidence. GATE 01 ### REACH Can AI read your information? 0 observed ✓ GATE 02 ### SHORTLIST Will it put you on the candidate list? 0 -5 · picked competitors, not you GATE 03 ### ACCURATE & RECOMMENDED Is your brand described accurately — and recommended? 0 -2 · mentioned-not-recommended / stale price / discontinued 03 · The Loop ## From spotting the gap to continuous improvement. The right order: see clearly first, then act; act on evidence; verify after acting. Edith OBSERVE ### Find the gap Pin down where the brand is absent, less competitive or misrepresented in AI answers. Homer ALIGN ### Build the case Establish an approved, verifiable factual basis. Friday ACT ### Close the gap Close information gaps one by one. Edith PROVE ### Retest What changed? Retest under the same conditions. Edith · Friday OPERATE ### Operate Turn the full loop into a repeatable capability. 04 · The Strategy ## Every confirmed gap gets a matching optimization playbook. Winin compares the question, model answers and visible sources to identify where the gap occurs, then turns that finding into executable work. L1 ### Fact governance Conflicts in identity, product facts, price and policy: first pin down source, owner and validity to form a usable version — before touching public content. → Homer fact charter L2 ### Site & product pages For confirmed gaps, prescribe fact pages, product pages, FAQs, comparison content and structured data. → Page & data tasks L3 ### Enterprise channels Locate the official stores, distributors, catalogs, partner profiles and product data that need correcting — as channel governance tasks. → Channel correction list L4 ### Independent sources Where independent evidence is missing, plan first-party research, test evidence, media materials, expert views or partner collateral. → External credibility plan 05 · The Products ## Find the gap. Govern the facts. Advance the work. Homer is the enterprise knowledge foundation: approved facts with their sources, versions and public permissions. Edith finds the gaps, Homer governs the facts, and Friday advances authorized optimization work. *Full intro →* ### [Continuously monitor how AI describes your brand.](/en/products/edith/) ** ### [One brand. One voice.](/en/products/homer/) ** ### [You approve. Friday advances authorized work.](/en/products/friday/) 06 · Services ## From diagnosis to execution, every step has a matching service. Measure & Diagnose ### AI recommendation diagnostics See where you’re considered, recommended, or absent — every answer keeps its raw content, reasons and cited sources. ### Managed monitoring & retesting Continuously track changes across models, competitors, facts and sources; same-conditions retesting keeps before-and-after changes on record. Knowledge Foundation ### Enterprise knowledge foundation building Unify brand info scattered across site, e-commerce, press and internal docs — single source of truth, owner, version control. ### Encode the foundation for AI Turn approved facts into structured public content that machines can parse and teams can trace; diagnosis and optimization share the same factual source. Execute & Verify ### Brand GEO Using facts approved in Homer, improve how the brand is understood, considered and recommended across site content, channels and external sources. ### Product GEO For category, comparison and buying questions, close the gaps in product facts and recommendation reasons. 07 · The Outcomes ## After the work, your brand gains four operating capabilities. These interfaces illustrate product capabilities. Formal results depend on retained answers, verified publication and comparable retesting. Answer evidence Answer record · archived Your brand Answer · reasons · sources Retain the answer; trace omissions, errors and source changes. Consideration Candidate position · monitored BrewMaster Aroma your brand ✓ Peak Track the questions where the brand enters consideration. Approved facts Approved facts · version 4 ✓ ERP source ✓ Marketing owner ✓ Valid to Dec 31 Source, owner, version and validity stay linked. Comparable change Recommendation retest · illustrative Retest after verified publication on the same basis. 08 · Client Cases ## Winin completed 14 tasks. You made the final decision. Three anonymized scenarios show how Winin moves from monitoring to approved action and comparable retesting. The following are anonymized client cases. Client names are not disclosed; the work, measurement scope and project outcomes are shown within the delivery record. 90 days 120 days 60 days Consumer brand Small appliances · Espresso machines 90 days ### From “AI doesn’t know you” to category first-candidate Starting point: 7 of 9 models never mentioned the brand in “espresso recommendations”; the flagship quoted at a stale price by 2. 11*%* → 37*%* Share of recommendation 14 wrong facts corrected +210*%* AI-referred inquiries Mon 09:12 approval required Tue 08:00 Awaiting confirmation Tue 14:30 review and authorization Wed 10:05 verified Fri 18:00 report Cross-border e-commerce Portable power stations 120 days ### Measuring from “mentioned” all the way to “bought” AI platforms already show product cards in answers. We build the full chain: how much drops at each cell, why, and how to fix it. Mentioned 61% Product card shown 34% Click-through to store 12% Attributed conversion 3.8% New-customer acquisition cost 38% below paid search · retestable on same basis B2B Industrial inspection SaaS 60 days ### Winning the reasons back in “vendor comparisons” Buyers ask AI first “which vendor is reliable”. Before: AI knew the name, but every recommendation reason went to competitors. 2 → 7 comparison-question reason coverage +64*%* demo bookings 0 → 6 models shortlisting you 09 · The Difference ## Pure monitoring tools stop at the first cell. Monitoring tools Winin monitors brand performance in AI answers ✓ ✓ 23 checks locate “why not recommended” — ✓ Friday automates preparation and advances work within the approved scope — ✓ same-conditions comparison before/after — ✓ subscription-based monitoring & iteration — ✓ Every number traces back to question · model · raw answer · time · source · denominator. Friday only uses facts approved in Homer, within validity, traceable to source. External evidence links to real research, media, experts or partners — no spam, no black-hat. Retest with identical questions and conditions; state what changed and what remains uncertain. growth · pipeline · sales conversion monitor → align → execute → retest → operate Facts checkable · Evidence traceable · Attribution restrained 10 · Pricing ## See the problem clearly, then decide how much to invest. ### Free Check ¥0 5 fixed questions · one sample per question · up to 3 active model APIs; an initial signal, not a formal baseline. ### Formal Baseline Scoped quote Custom questions, models and cadence; a complete retestable baseline. CORE ### Brand · Product GEO Scoped quote Optimization strategy, site/content/channel execution, and same-conditions retesting. ### Continuous Monitoring Subscription Ongoing tracking of AI recommendation shifts, competitor moves and new growth openings. 11 · FAQ ## First the concepts, then the work. The most-asked questions about GEO and AEO — including the uncomfortable ones. + GEO (Generative Engine Optimization) improves the conditions for accurate interpretation, consideration and citation through approved facts and publicly verifiable expressions; measured change still requires retesting. SEO focuses on discovery and clicks in search results, while GEO focuses on how generative answers understand and represent you. + AEO (Answer Engine Optimization) focuses on making direct answers easy to retrieve and support. GEO overlaps with AEO but also examines how generative systems understand, compare, consider and recommend a brand or product. Both depend on clear facts, accessible content and comparable measurement. + Yes. SEO improves how content is crawled, indexed and discovered in search; GEO examines how generative systems understand and use that information. Clear fact pages, product information and structured data can support both, but search performance and changes in AI answers should be measured separately. + AI may draw on training data, third-party media, marketplace pages, user content and your site. Homer organizes scattered information into approved facts with clear, traceable expressions; whether a model uses them, and how it responds, still has to be observed. + Training data, retrieval paths and generation strategies are all updated continuously, so the same question can get different answers on different days — that is normal. What matters is having a stable baseline to measure the direction of change, not drawing conclusions from a single answer. + Structured data is a set of machine-readable annotations (such as Schema or JSON-LD) that states what something is, its price and whether it is available. It makes explicit facts easier to identify; whether AI or search systems use them, and how they express them, still has to be observed. ## Source This Markdown document is generated from the same approved page source as https://winin.ai/; the canonical HTML page is the public presentation version.