How automated tools fit into Boba.guide’s editorial work
AI Use Policy
Boba.guide may use artificial intelligence and other automated tools to assist parts of research, writing, editing, translation, data organization, image production and website development. AI does not replace the evidence behind a factual claim or the editorial responsibility for what we publish.
That distinction is fundamental. A language model can help identify questions, compare material, reorganize information or expose gaps worth researching. It can also produce a confident answer that is incomplete, outdated, incorrectly attributed or entirely fabricated.
For that reason, Boba.guide treats AI output as working material — not as an independent source of truth.
The central rule
AI Can Assist the Work. It Cannot Carry the Evidence.
An AI system may help organize a question about boba, suggest angles to investigate, compare long documents, surface possible contradictions, draft language or assist with technical production.
But if a published sentence says a drink contains a particular ingredient, a Taiwanese term means something specific, a shop charges a certain price, a topping has a particular nutritional value or a safety recommendation applies, the support for that claim must come from appropriate evidence — not from the AI model saying so.
Editorial responsibility remains with Boba.guide.
Where AI may enter the workflow
Tasks AI May Assist With — and the Limits That Follow
Different uses create different risks. The fact that AI is acceptable for one task does not make every use equally appropriate.
Finding questions and research directions
AI may help brainstorm terminology, identify possible subtopics, compare broad themes across existing material or suggest questions that deserve further investigation.
Suggested facts, citations and conclusions must still be checked independently.
Summarizing or comparing source material
Automated tools may help navigate long menus, reports, regulations, research papers or source collections. Important claims are checked against the underlying source rather than accepted solely from an automated summary.
Producing or restructuring working copy
AI may assist with early drafts, rewrites, organization, headings, comparisons and explanations. Published copy still has to meet Boba.guide’s standards for factual support, scope, usefulness and uncertainty.
Helping interpret material in other languages
Machine-assisted translation can be useful when researching Taiwanese, Japanese, Korean, Chinese or other local-language sources. Important terminology is checked in context because literal translation can erase distinctions that matter.
Grammar, clarity and consistency
AI-assisted tools may help identify awkward wording, repetition, formatting issues or structural inconsistencies. Editorial polishing does not change the requirement that substantive claims remain supported.
Sorting, calculating or comparing structured information
Automation may help organize menus, prices, measurements, ingredient data or other structured material. Calculations, units, source rows and important transformations should be checked before publication.
Code, layout and technical assistance
AI tools may assist with HTML, CSS, responsive layouts, accessibility checks, debugging and other production tasks. Generated code is not assumed to be correct merely because it renders.
Creating facts, citations, credentials or experiences
AI must not be used to manufacture a source, invent a shop visit, create a quotation, fabricate a credential, claim a test that did not happen or generate a number that is then presented as researched fact.
A critical distinction
AI Is Not a Source Citation
If an AI system tells us that a tapioca pearl traditionally contains a particular ingredient, that is a research prompt — not the evidence we cite internally for the claim.
If it says a tea term has a specific meaning in Taiwan, we should check local-language usage, menus, tea sources or other appropriate evidence.
If it supplies a current price, nutrition number, caffeine value, ingredient list or legal requirement, those details require verification from the relevant current source.
An AI response says: “50% sweetness means half the sugar of the finished drink.”
That statement cannot simply be published. A shop’s percentage may control only one adjustable sweetener while pearls, fruit syrups, powders or other ingredients contribute additional sugar. The actual meaning must be researched in context.
The verification gate
What Must Be Checked Beyond the AI Output
People, organizations, shops, products and locations should not be created or attributed solely because a model produced a plausible name.
Prices, serving sizes, percentages, temperatures, calories, sugar, caffeine, ratios, dates and measurements need a traceable basis appropriate to the claim.
A model-generated quotation is not a real quotation. Quotes must come from a real identifiable source.
AI memory or search summaries cannot establish what a chain or shop currently sells. Current commercial facts should be checked against current evidence.
Repeated origin stories need source review, particularly when credible competing claims exist.
Allergens, food safety, choking, caffeine, pregnancy, diabetes, dietary suitability and similar claims require stronger evidence than generated text.
AI can broaden research, but it can also flatten culture
Using AI in Multilingual and International Boba Research
Boba.guide deliberately researches beyond English-language sources. AI can make that work faster, but speed creates a particular risk: turning culturally specific terms into convenient but inaccurate English equivalents.
AI may help identify possible Traditional Chinese, Japanese, Korean or other search terms. Those terms should then be used to find actual local sources rather than treating the generated translation itself as proof.
A machine translation may tell us the general direction of a drink name. When the distinction affects ingredients, tea type, texture or cultural meaning, the wording requires contextual review.
Terms such as 珍珠, 波霸, 粉圓, 鮮奶茶 or 奶蓋 should not automatically be collapsed into a single generic English term if their use differs by market or context.
AI-generated summaries can reproduce stereotypes or oversimplified origin stories. Taiwan and other mature boba cultures should be researched as living food and tea cultures, not reduced to decorative clichés.
Responsibility does not move to the software
What Human Editorial Oversight Means at Boba.guide
Decide what the page should actually answer
AI can generate endless adjacent topics. Editorial judgment determines what belongs on the page and what should remain outside its scope.
Connect factual statements to appropriate evidence
The supporting authority should come from the underlying source: a regulator, research paper, official menu, ingredient disclosure, original document or other suitable evidence.
Decide how strong the wording can responsibly be
A model may confidently call something “the best,” “the most popular” or “traditional.” Editors need to determine whether the evidence supports that level of certainty.
Make the finished page useful rather than merely fluent
Good prose is not enough. The material needs useful distinctions, examples, caveats, measurements, comparisons or decisions appropriate to the reader’s question.
Take responsibility for what remains published
If AI-assisted work contributes to an error, “the AI made it up” is not an acceptable correction policy. The published error belongs to Boba.guide and should be handled accordingly.
Higher stakes require higher scrutiny
AI Assistance Does Not Lower the Evidence Standard
AI may assist with organizing flavor descriptions or preference scenarios, but subjective recommendations should remain framed as recommendations rather than universal facts.
Exact formulations vary. Generated ingredient lists should not be substituted for shop-specific or manufacturer information where the recipe matters.
Calories, sugar, caffeine and serving data require actual supporting data and appropriate serving context.
AI should not infer that a drink is dairy-free, gluten-free, vegan or allergen-free merely from a product name, appearance or generic recipe.
Food-safety and health-related guidance must be grounded in stronger authoritative evidence. Generative output does not become medical or safety authority.
AI must not invent quotations, affiliations, ownership, professional credentials, store details or claims about a person’s conduct.
Visual content needs its own rules
AI-Generated and AI-Assisted Images
Boba.guide may use generated or AI-assisted visual material where appropriate to illustrate drinks, ingredients, concepts or editorial themes.
Generated imagery must not be used as documentary proof
A generated cup cannot prove what a specific restaurant serves, how a real shop looks, what a commercial product’s packaging contains or what happened at a historical event.
Images should not manufacture brand evidence
We do not intentionally create fake logos, menus, storefronts or branded products and then present them as photographs of real businesses.
Food should remain plausible
Generated beverage imagery should avoid impossible pearl geometry, misleading ingredients, duplicated cups, fake menu text or other artifacts that materially misrepresent the subject.
Real supplied photography has separate requirements
Ownership, permission, attribution and the distinction between documentary and illustrative material are covered in our Image and Photography Policy.
Not everything belongs in an AI prompt
Privacy, Confidential Material and Reader Information
Using an AI tool does not remove the responsibility to think about what information is being supplied to it.
Account passwords, authentication secrets, private access tokens and similar security information should not be entered into generative tools for editorial convenience.
Personal information supplied through correspondence should not be unnecessarily exposed to automated systems merely because it could make drafting faster.
Private material provided under a legitimate expectation of confidentiality requires appropriate handling regardless of whether an AI tool is available.
When automation is useful, the practical principle is to use only the information necessary for the task rather than copying an entire private record into a tool.
Transparency should match materiality
When AI Use Should Be Visible to Readers
Not every spellcheck, code suggestion or internal research query needs a large disclosure banner. What matters is whether AI materially changes the nature of what the reader is seeing or needs to know to interpret it correctly.
Grammar checking, formatting assistance, code suggestions, internal outlining or similar behind-the-scenes uses do not normally require a prominent article-level disclosure on their own.
If a published product depends substantially on automated translation, summarization or another transformation where that fact materially affects interpretation, disclosure may be appropriate.
AI-generated material must not be allowed to mislead a reader into believing that a synthetic depiction is documentary evidence of a real person, shop, event or product.
If the fact that something was generated or produced using AI is central to the story or demonstration, that context should be made clear rather than hidden.
Hard boundaries
What We Do Not Use AI to Pretend Happened
We do not invent first-hand shop visits, tastings, interviews, kitchen tests or experiments that did not occur.
We do not fabricate quotations and attribute them to real people.
We do not create fictional experts, contributors, professional credentials or reviewer biographies to make a page look more authoritative.
We do not invent citations, official documents, scientific papers, menu items, prices or nutrition data.
We do not treat AI-generated consensus as evidence that a practice is universal across boba shops or countries.
We do not deliberately use synthetic visual material to impersonate documentary photography of a real shop, product or event.
We do not use automation as an excuse to leave a known factual error uncorrected.
AI-assisted work can still be wrong
Corrections Apply Regardless of How the Error Was Produced
If incorrect information reaches a published Boba.guide page, the responsibility does not disappear because an automated tool contributed to the mistake.
We evaluate the underlying error, correct or clarify it as appropriate and use the same editorial standard that applies to other published material.
The wider editorial framework
Policies That Work Together With This One
AI use is one part of a broader editorial system. The standards for evidence, contributors and accountability continue to apply whether automation is involved or not.
How Boba.guide weighs evidence, handles uncertainty and researches boba across languages and markets.
Why published responsibility is not replaced by a fictional expert persona or an automated system.
Questions about this policy or another part of the site’s editorial work can be sent through the main contact route.
Questions about AI use at Boba.guide
Contact the Editorial Team
If you believe automated or AI-assisted material on Boba.guide has introduced a factual problem, misleading image, attribution issue or another concern, tell us specifically what you found.
Include the page URL and enough detail for us to locate and review the issue.
info@boba.guide