[{"data":1,"prerenderedAt":16},["ShallowReactive",2],{"blog-post-prioritize-chatbot-sources":3},{"id":4,"slug":5,"title":6,"metaDescription":7,"heroImageUrl":8,"languageCode":9,"publishedAt":10,"updatedAt":11,"contentHtml":12,"faqJsonLd":13,"keywords":14},1013621,"prioritize-chatbot-sources","How to Prioritize Chatbot Sources for Accurate Answers","Learn how to rank FAQs, web pages and product catalogs as chatbot sources, handle conflicts and gaps, test answers and know when to hand off to a human.","https:\u002F\u002Fmedia.babylovegrowth.ai\u002Fblog-images\u002Forganization-42511\u002F1792856890672_Owner-comparing-chatbot-knowledge-sources.jpeg","en","2026-10-24T15:48:57.126Z","2026-10-24T16:04:24.819Z","\u003Cp>To prioritize chatbot sources, rank them by authority and by how fast they change. Put your policies and structured data first, then curated FAQs, then general website content, and let the chatbot admit a gap instead of guessing. Give each source an owner, a way to refresh, and a test that proves the chatbot uses it correctly.\u003C\u002Fp>\n\u003Cp>This guide shows how to do that on a small business website: what to rank first, what to do when sources disagree, how to test, when to hand off to a person, and how to adapt it all for several languages.\u003C\u002Fp>\n\u003Cdiv data-blg-cta=\"after_tldr\" data-blg-cta-layout=\"strip\" style=\"margin:28px 0;font-family:-apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif;box-sizing:border-box;\">\u003Cdiv style=\"border-radius:26px;padding:min(14px,3.2vw);background:radial-gradient(circle at 100% 0%,#d7f2fc 0 150px,rgba(255,255,255,0) 151px),radial-gradient(circle at 0% 100%,#d7f2fc 0 130px,rgba(255,255,255,0) 131px),linear-gradient(180deg,#e4f6fd 0%,#f2fbfe 100%);\">\u003Cdiv style=\"background:#ffffff;border-radius:18px;box-shadow:0 12px 34px rgba(0,0,0,0.13);overflow:hidden;\">\u003Cdiv style=\"display:flex;flex-wrap:wrap;align-items:center;gap:16px 22px;padding:20px 24px;\">\u003Cdiv style=\"flex:1 1 260px;min-width:0;\">\u003Cdiv style=\"margin:0 0 8px;\">\u003Cspan style=\"display:inline-block;max-width:100%;box-sizing:border-box;border-radius:999px;padding:6px 13px;font-size:12px;font-weight:800;letter-spacing:0.1em;text-transform:uppercase;line-height:1.3;background:#1FB8EF;color:#ffffff;\">Konvuno\u003C\u002Fspan>\u003C\u002Fdiv>\u003Cdiv style=\"font-size:19px;font-weight:800;line-height:1.2;letter-spacing:-0.01em;color:#1f2937;margin:0;\">Keep Chatbot Answers Grounded\u003C\u002Fdiv>\u003Cdiv style=\"font-size:14px;line-height:1.5;color:#64748b;margin-top:4px;\">Konvuno answers from your FAQs, website content and product catalog, with product feeds syncing automatically to keep product details current.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv style=\"flex:0 0 auto;\">\u003Ca href=\"https:\u002F\u002Fkonvuno.com\u002F\" style=\"display:inline-flex;align-items:center;gap:9px;border-radius:10px;font-weight:700;font-size:15px;text-decoration:none;padding:13px 22px 13px 26px;background:#1FB8EF;color:#ffffff;\">Explore Konvuno\u003Csvg width=\"17\" height=\"17\" viewBox=\"0 0 24 24\" fill=\"none\" stroke=\"#ffffff\" stroke-width=\"2.6\" stroke-linecap=\"round\" stroke-linejoin=\"round\" style=\"display:block;flex:0 0 auto;\">\u003Cpath d=\"M5 12h14\"\u002F>\u003Cpath d=\"m12 5 7 7-7 7\"\u002F>\u003C\u002Fsvg>\u003C\u002Fa>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\n\u003Ch2 id=\"tldr\" tabindex=\"-1\">TL;DR\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>Rank sources by authority: policies and live product data first, curated FAQs second, general page copy third, old blog posts last.\u003C\u002Fli>\n\u003Cli>Retrieval quality matters as much as the content itself. A well-grounded answer can still be wrong if the chatbot pulls the wrong passage.\u003C\u002Fli>\n\u003Cli>Write down what happens when sources conflict or go missing: the chatbot should say it is unsure and route the question to a person, not improvise.\u003C\u002Fli>\n\u003Cli>Test with real customer questions, including edge cases such as discontinued products, before you launch and after every big change.\u003C\u002Fli>\n\u003Cli>Treat each language as its own source set with its own review, not as an automatic translation of the English one.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 id=\"table-of-contents\" tabindex=\"-1\">Table of Contents\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"#what-your-chatbot-should-answer-and-from-what\">What your chatbot should answer, and from what\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#why-retrieval-is-the-control-point\">Why retrieval is the control point\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#a-practical-priority-order\">A practical priority order\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#what-to-do-when-sources-conflict-run-out-or-go-stale\">What to do when sources conflict, run out or go stale\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#keeping-answers-accurate-as-things-change\">Keeping answers accurate as things change\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#testing-answers-and-finding-gaps-in-your-sources\">Testing answers and finding gaps in your sources\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#when-the-chatbot-should-hand-off-to-a-person\">When the chatbot should hand off to a person\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#adapting-sources-for-multilingual-visitors\">Adapting sources for multilingual visitors\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#our-perspective-start-small-curate-hard\">Our perspective: start small, curate hard\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#putting-it-into-practice-with-konvuno\">Putting it into practice with Konvuno\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#faq\">FAQ\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"#sources\">Sources\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 id=\"what-your-chatbot-should-answer-and-from-what\" tabindex=\"-1\">What your chatbot should answer, and from what\u003C\u002Fh2>\n\u003Cp>A website chatbot earns its place by answering the questions your visitors ask over and over: what you sell, what it costs, whether it is in stock, how shipping and returns work, how to book, and who to contact. It should not be expected to answer everything. Questions about a specific order, a refund exception or a legal matter belong with a person.\u003C\u002Fp>\n\u003Cp>For a small business, the information a chatbot can draw on usually falls into four groups:\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cstrong>Structured data\u003C\u002Fstrong>, such as a product catalog with prices and availability.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Curated answers\u003C\u002Fstrong>, such as an FAQ you wrote on purpose.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Website content\u003C\u002Fstrong>, such as service pages, about pages and blog posts.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Policies and documents\u003C\u002Fstrong>, such as shipping terms, return rules and warranty text.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Cp>These groups are not equal. A price in your live catalog is more trustworthy than a price mentioned in a blog post from three years ago. A sentence in your returns policy outranks a casual line in a product description. Prioritizing sources means making those differences explicit before the chatbot has to choose between them.\u003C\u002Fp>\n\u003Ch2 id=\"why-retrieval-is-the-control-point\" tabindex=\"-1\">Why retrieval is the control point\u003C\u002Fh2>\n\u003Cp>Most modern website assistants use retrieval-augmented generation: the system searches your content for relevant passages, then writes an answer based on what it found. Google Cloud’s explanation of \u003Ca href=\"https:\u002F\u002Fcloud.google.com\u002Fuse-cases\u002Fretrieval-augmented-generation\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">retrieval-augmented generation\u003C\u002Fa> makes a point that matters for owners: retrieval is the critical control point. Even a well-grounded answer can be irrelevant or wrong if the search step brings back poor passages.\u003C\u002Fp>\n\u003Cp>In practice, that means two jobs, not one:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Curate the content.\u003C\u002Fstrong> Remove what is outdated, duplicated or contradictory.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Shape what gets found.\u003C\u002Fstrong> Clear titles, one topic per page and plain wording help the right passage surface for the right question.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>The second job overlaps with how content gets found by AI search tools generally. If you want a deeper look at how pages become findable by AI systems, this piece on \u003Ca href=\"https:\u002F\u002Fphenyx.co\u002Fpost\u002Fhow-to-show-up-in-chatgpt-and-ai-overviews\" target=\"_blank\" rel=\"noopener\">making content accessible to AI\u003C\u002Fa> covers the access side, and a walkthrough of \u003Ca href=\"https:\u002F\u002Fbabylovegrowth.ai\u002Fblog\u002Frank-in-chatgpt\" target=\"_blank\" rel=\"noopener\">fast retrieval actions\u003C\u002Fa> shows how retrieval-friendly writing works from the outside. The same habits, such as one clear answer per page and descriptive headings, also help your own chatbot.\u003C\u002Fp>\n\u003Ch2 id=\"a-practical-priority-order\" tabindex=\"-1\">A practical priority order\u003C\u002Fh2>\n\u003Cp>Here is a ranking that works for most small shops and service businesses. Adjust it to your situation, but decide it in writing.\u003C\u002Fp>\n\u003Cdiv class=\"kv-doc-table\">\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Priority\u003C\u002Fth>\n\u003Cth>Source type\u003C\u002Fth>\n\u003Cth>Why it ranks here\u003C\u002Fth>\n\u003Cth>Typical owner\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>1\u003C\u002Ftd>\n\u003Ctd>Live product catalog (prices, stock, variants)\u003C\u002Ftd>\n\u003Ctd>Changes often, and wrong data costs sales\u003C\u002Ftd>\n\u003Ctd>Store manager\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>2\u003C\u002Ftd>\n\u003Ctd>Policies (shipping, returns, warranty, privacy)\u003C\u002Ftd>\n\u003Ctd>Carry legal and financial weight\u003C\u002Ftd>\n\u003Ctd>Owner or operations\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>3\u003C\u002Ftd>\n\u003Ctd>Curated FAQ\u003C\u002Ftd>\n\u003Ctd>Written to answer exact questions in your voice\u003C\u002Ftd>\n\u003Ctd>Support lead or owner\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>4\u003C\u002Ftd>\n\u003Ctd>Service and product pages\u003C\u002Ftd>\n\u003Ctd>Broad, accurate when current, but not written as answers\u003C\u002Ftd>\n\u003Ctd>Marketing or web person\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>5\u003C\u002Ftd>\n\u003Ctd>Blog posts and news\u003C\u002Ftd>\n\u003Ctd>Dated by nature, useful for context only\u003C\u002Ftd>\n\u003Ctd>Marketing\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>6\u003C\u002Ftd>\n\u003Ctd>Old PDFs, archived pages\u003C\u002Ftd>\n\u003Ctd>Highest risk of stale facts\u003C\u002Ftd>\n\u003Ctd>Remove or archive\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch3 id=\"put-structured-data-first-for-facts-that-change\" tabindex=\"-1\">Put structured data first for facts that change\u003C\u002Fh3>\n\u003Cp>Prices, availability and variants should come from a feed, not from copy someone typed into a page. If you run WooCommerce, or publish a Google Merchant or Facebook product feed, the catalog is already your single source of truth for these facts. A chatbot that reads the feed will give the same price your checkout does. A chatbot that reads a page where someone typed the price by hand will drift out of date the first time you run a sale.\u003C\u002Fp>\n\u003Ch3 id=\"write-the-faq-for-the-chatbot-not-just-for-people\" tabindex=\"-1\">Write the FAQ for the chatbot, not just for people\u003C\u002Fh3>\n\u003Cp>A curated FAQ is the highest-leverage source you control. Each entry should answer one question completely, in the words customers actually use. Add questions as they arise. The best source for new entries is your own conversation log: the questions the chatbot could not answer last week are the FAQ entries to write this week.\u003C\u002Fp>\n\u003Ch3 id=\"treat-general-pages-as-supporting-evidence\" tabindex=\"-1\">Treat general pages as supporting evidence\u003C\u002Fh3>\n\u003Cp>Service pages and about pages are written to persuade, not to answer. They can be vague (“fast delivery”) where a customer needs a number. Keep them in the knowledge base, but rank them below the FAQ and policies so that a precise answer wins over a marketing phrase.\u003C\u002Fp>\n\u003Ch3 id=\"retire-what-you-do-not-want-quoted\" tabindex=\"-1\">Retire what you do not want quoted\u003C\u002Fh3>\n\u003Cp>If a page should not be used as an answer, take it out of the chatbot’s reach. A 2021 blog post about a discontinued product line is a liability, not an asset. Archiving it from the knowledge base costs nothing and removes a whole class of wrong answers.\u003C\u002Fp>\n\u003Ch2 id=\"what-to-do-when-sources-conflict-run-out-or-go-stale\" tabindex=\"-1\">What to do when sources conflict, run out or go stale\u003C\u002Fh2>\n\u003Cp>Every site has contradictions: the footer says returns take 30 days, an old FAQ says 14. The question is not whether this happens but what the chatbot does when it does.\u003C\u002Fp>\n\u003Cp>ALESO’s guidance on \u003Ca href=\"https:\u002F\u002Fwww.aleso.ch\u002Fen\u002Finsights\u002Fknowledge-data-rag\u002Fconflicting-outdated-missing-sources\u002F\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">conflicting, outdated or missing sources\u003C\u002Fa> recommends defining authority and validity rules, logging conflicts together with the query, results and versions, exposing unresolved contradictions, and routing them to the person responsible. You can apply that on a small scale without any engineering:\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cstrong>Set an authority rule.\u003C\u002Fstrong> For example: the live catalog beats any page, policies beat the FAQ, and the FAQ beats general copy.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Set a validity rule.\u003C\u002Fstrong> Anything with a date, such as a promotion or seasonal hours, needs an end date or a reviewer.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Decide what a gap looks like.\u003C\u002Fstrong> If no source answers the question, the chatbot should say so plainly and offer a human, never fill the silence with a plausible guess.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Name an owner for each conflict.\u003C\u002Fstrong> A flagged contradiction nobody owns will still be there next quarter.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Fix the source, not the symptom.\u003C\u002Fstrong> If the chatbot gave a wrong answer because two pages disagreed, correct the pages. Patching one answer leaves the conflict in place.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Ch3 id=\"missing-information-is-a-feature-of-the-dashboard-not-a-failure\" tabindex=\"-1\">Missing information is a feature of the dashboard, not a failure\u003C\u002Fh3>\n\u003Cp>A question the chatbot cannot answer is a signal. Review unanswered questions weekly and sort them into three piles: write a new FAQ entry, fix an existing page, or leave it for a human because it should never be automated. Over a few months the unanswered list gets shorter and the content gets better.\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Fmedia.babylovegrowth.ai\u002Fblog-images\u002Forganization-42511\u002F1792856933340_Three-paths-for-unanswered-chatbot-questions.jpeg\" alt=\"Three paths for unanswered chatbot questions\">\u003C\u002Fp>\n\u003Ch2 id=\"keeping-answers-accurate-as-things-change\" tabindex=\"-1\">Keeping answers accurate as things change\u003C\u002Fh2>\n\u003Cp>Accuracy decays. Prices move, services are added, hours change for the holidays. The way to keep up is to remove as many manual steps as you can and put a calendar on the rest.\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Automate what is structured.\u003C\u002Fstrong> Catalog sync from WooCommerce or a product feed keeps prices and availability current without anyone touching the chatbot.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Review what is written.\u003C\u002Fstrong> Put FAQ and policy reviews on a recurring schedule, such as monthly for small shops and before every seasonal campaign.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Tie content changes to chatbot changes.\u003C\u002Fstrong> Any time you update a policy page, update the matching FAQ entry in the same sitting.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Watch the questions.\u003C\u002Fstrong> A sudden jump in one question usually means something changed: a shipping delay, a viral product, a confusing page.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>A useful habit is to treat a launch like a release: before you publish a new service or run a sale, ask what the chatbot will say about it tomorrow.\u003C\u002Fp>\n\u003Ch2 id=\"testing-answers-and-finding-gaps-in-your-sources\" tabindex=\"-1\">Testing answers and finding gaps in your sources\u003C\u002Fh2>\n\u003Cp>Do not trust a chatbot because the demo looked good. Build a test set from reality. Swept’s \u003Ca href=\"https:\u002F\u002Fwww.swept.ai\u002Fpost\u002Fai-customer-service-agent-evaluation-framework\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">framework for evaluating AI customer service agents\u003C\u002Fa> suggests testing real customer questions, edge cases and adversarial inputs, including discontinued products and regional exceptions, and judging correctness, policy adherence, factual grounding and escalation quality. It also advises building your own test suite before deployment.\u003C\u002Fp>\n\u003Cp>A workable version for a small business:\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cstrong>Collect 30 to 50 real questions\u003C\u002Fstrong> from email, chat history, phone notes and social messages.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Add the awkward ones.\u003C\u002Fstrong> Ask about a product you stopped selling, a country you do not ship to, a discount that ended, and a question that mixes two topics.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Try to break it.\u003C\u002Fstrong> Ask it to ignore its instructions, to promise a refund, or to comment on a competitor. A good assistant stays within its sources and scope.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Score each answer\u003C\u002Fstrong> on four points: is it correct, does it follow your policy, can you point to the source it used, and did it hand off when it should have?\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Trace every failure to a source.\u003C\u002Fstrong> Wrong answer, missing page, conflicting pages or poor retrieval? Each has a different fix.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Repeat after big changes\u003C\u002Fstrong>, such as a new catalog, a redesigned site or a policy update.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Cp>The sixth step is the one most people skip. A test set you run once is a launch checklist. A test set you rerun is quality control.\u003C\u002Fp>\n\u003Ch2 id=\"when-the-chatbot-should-hand-off-to-a-person\" tabindex=\"-1\">When the chatbot should hand off to a person\u003C\u002Fh2>\n\u003Cp>Escalation is part of accuracy. Swept’s framework argues that a system has to recognize the limits of its own capability, notice frustration and high-consequence topics, and pass useful context to the right human queue. That is a sound way to define handoff rules for a small team:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>The question is outside the sources.\u003C\u002Fstrong> No confident answer exists, so a person should take it.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The stakes are high.\u003C\u002Fstrong> Refund disputes, safety concerns, legal or medical questions and complaints should reach a human quickly.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The visitor is frustrated.\u003C\u002Fstrong> Repeated rephrasing or an explicit request for a person is the cue.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The visitor is ready to buy or book.\u003C\u002Fstrong> A warm lead with a custom request is worth a human reply.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>How the handoff works matters as much as when. The visitor should not have to repeat themselves, so the person who picks up needs the conversation so far and any contact details already captured. If nobody is online, the chatbot should say when to expect a reply and collect an email address rather than leaving the visitor at a dead end.\u003C\u002Fp>\n\u003Ch2 id=\"adapting-sources-for-multilingual-visitors\" tabindex=\"-1\">Adapting sources for multilingual visitors\u003C\u002Fh2>\n\u003Cp>If your audience speaks more than one language, the source question multiplies. A few rules keep it manageable:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Decide the source of truth.\u003C\u002Fstrong> Usually one language holds the master version of policies, and the others follow it.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Do not rely on word-for-word translation.\u003C\u002Fstrong> Shipping terms, units, currency, date formats and legal wording differ by market. Write local FAQ entries where the answer is genuinely different.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Keep catalogs consistent.\u003C\u002Fstrong> If your feed has translated titles and descriptions, the chatbot can answer in the visitor’s language from the same product data.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Test in each language.\u003C\u002Fstrong> Run your question set separately. A gap that does not exist in English often shows up in the second language.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Check handoff coverage.\u003C\u002Fstrong> If the chatbot speaks Spanish but your staff does not, set expectations about reply times and languages.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 id=\"our-perspective-start-small-curate-hard\" tabindex=\"-1\">Our perspective: start small, curate hard\u003C\u002Fh2>\n\u003Cp>We build Konvuno for owners who have a website and no support team, and what we see is consistent: the businesses with the best chatbot results are not the ones with the most content, they are the ones with the most deliberate content. A short, accurate FAQ and a live product feed beat a hundred loosely maintained pages.\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Fmedia.babylovegrowth.ai\u002Fblog-images\u002Forganization-42511\u002F1792856918736_Our-perspective-start-small-curate-hard-overview-diagram.jpeg\" alt=\"Our perspective: start small, curate hard — overview diagram\">\u003C\u002Fp>\n\u003Cp>Our advice is to begin with the sources that answer most visitor questions, connect your catalog so prices and stock take care of themselves, and read the analytics on what visitors actually ask. Let the questions tell you what to write next. Configure everything, preview the answers, and only then go live. A chatbot should feel like your best-informed employee, one who says “let me get a person” when the answer is not in the file.\u003C\u002Fp>\n\u003Cp>No developer should be needed for any of this. The point of a hosted dashboard is that the person who knows the business can change the sources directly.\u003C\u002Fp>\n\u003Cblockquote>\n\u003Cp>\u003Cem>— Konstantin\u003C\u002Fem>\u003C\u002Fp>\n\u003C\u002Fblockquote>\n\u003Ch2 id=\"putting-it-into-practice-with-konvuno\" tabindex=\"-1\">Putting it into practice with Konvuno\u003C\u002Fh2>\n\u003Cp>Konvuno answers visitor questions from your FAQs, website content and product catalog, captures leads into a built-in CRM and hands off to a human when a question needs one. Catalogs sync automatically from WooCommerce or any Google Merchant or Facebook product feed, so answers about products, prices and availability stay current. It installs with a single script tag, speaks seven languages, and you can configure and preview everything before you pay.\u003C\u002Fp>\n\u003Cp>Plans are \u003Ca href=\"https:\u002F\u002Fkonvuno.com\u002F\" target=\"_blank\" rel=\"noopener\">Starter at €29 per month, Growth at €79 per month and Business at €199 per month\u003C\u002Fa>, with no per-seat pricing.\u003C\u002Fp>\n\u003Ch2 id=\"faq\" tabindex=\"-1\">FAQ\u003C\u002Fh2>\n\u003Ch3 id=\"which-source-should-a-chatbot-trust-first\" tabindex=\"-1\">Which source should a chatbot trust first?\u003C\u002Fh3>\n\u003Cp>Trust structured, frequently updated data first, which usually means your live product catalog and your written policies. Next comes a curated FAQ, then general page copy. Write this order down so the chatbot and your team resolve conflicts the same way.\u003C\u002Fp>\n\u003Ch3 id=\"how-often-should-i-update-my-chatbots-sources\" tabindex=\"-1\">How often should I update my chatbot’s sources?\u003C\u002Fh3>\n\u003Cp>Automate what changes fastest, such as prices and stock, through a catalog sync. Review FAQs and policies on a set schedule, for example monthly, and again before any sale or policy change. Check unanswered questions weekly, because they show you where the content has fallen behind.\u003C\u002Fp>\n\u003Ch3 id=\"what-should-a-chatbot-say-when-it-does-not-know-the-answer\" tabindex=\"-1\">What should a chatbot say when it does not know the answer?\u003C\u002Fh3>\n\u003Cp>It should say it does not have a reliable answer and offer a way forward: connect the visitor with a person, or take their contact details so someone can reply. Guessing is worse than admitting a gap, because a confident wrong answer about price or policy can cost you a sale or a dispute.\u003C\u002Fp>\n\u003Ch3 id=\"how-do-i-test-whether-my-chatbot-is-using-the-right-sources\" tabindex=\"-1\">How do I test whether my chatbot is using the right sources?\u003C\u002Fh3>\n\u003Cp>Build a list of real customer questions plus edge cases, such as discontinued products and regional exceptions. Score each answer for correctness, policy adherence, grounding in a source and quality of handoff. Rerun the list after any major content change.\u003C\u002Fp>\n\u003Ch3 id=\"do-i-need-separate-sources-for-each-language\" tabindex=\"-1\">Do I need separate sources for each language?\u003C\u002Fh3>\n\u003Cp>Not always, but you do need to check each language separately. Shared product data can serve all languages, while policies, shipping details and local rules often need their own wording. Test every language with its own set of questions.\u003C\u002Fp>\n\u003Ch2 id=\"sources\" tabindex=\"-1\">Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fcloud.google.com\u002Fuse-cases\u002Fretrieval-augmented-generation\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">What is Retrieval-Augmented Generation (RAG)? | Google Cloud\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.aleso.ch\u002Fen\u002Finsights\u002Fknowledge-data-rag\u002Fconflicting-outdated-missing-sources\u002F\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">What happens when sources conflict, are outdated or are missing | ALESO\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.swept.ai\u002Fpost\u002Fai-customer-service-agent-evaluation-framework\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">How to Evaluate AI Customer Service Agents | Swept\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fphenyx.co\u002Fpost\u002Fhow-to-show-up-in-chatgpt-and-ai-overviews\" target=\"_blank\" rel=\"noopener\">ChatGPT and AI Overviews: Marketers, Check Access First | PHENYX\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fbabylovegrowth.ai\u002Fblog\u002Frank-in-chatgpt\" target=\"_blank\" rel=\"noopener\">8 Fast Retrieval Actions to Rank in ChatGPT for Marketers\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>",null,[15],"prioritize chatbot sources",1794110722194]