Mandatory newsroom policy. Apply these rules before requesting or generating AI-assisted content, during editing, and again before publication. They support reader trust, Google Publisher Policy compliance and search quality; they do not guarantee AdSense approval. Google can update its policies, so editors must also consult the linked official policy pages.
1. Gate before AI generation
- Start with a real reporting brief: topic, intended audience, primary sources, confirmed facts, disputed claims, missing evidence and the original value HeadlineDecoded will add.
- Do not ask AI to fill factual gaps. It must label uncertainty, preserve attribution and avoid inventing names, dates, numbers, quotations, sources, links or image provenance.
- AI output is a draft, never a publication decision. The generation workflow must verify and substantially review all text, metadata, translations and generated imagery; unresolved checks must block publication.
- Do not mass-produce pages primarily to manipulate search rankings. Automated content must provide original reporting, analysis, synthesis or another clear reader benefit.
2. Accuracy, sourcing and transparency
- Every published story must use a verified public author or an explicitly identified collaborative newsroom desk. Never invent a journalist, biography, portrait, credential, location or social-media account to manufacture trust.
- Author pages must disclose the person’s real role, relevant expertise, genuine biography and only links that the newsroom has verified. Placeholder author records remain private and cannot be used for publication.
- Verify names, titles, dates, figures, quotations and material claims against primary documents or multiple credible sources.
- Distinguish confirmed fact, allegation, opinion, analysis and inference. Give subjects a fair opportunity to respond when allegations could harm them.
- Link or name important sources. Never disguise sponsored, affiliate or promotional material as independent reporting.
- Use a visible correction note for material post-publication changes.
- Generated or illustrative imagery must not be presented as documentary evidence. Label it when a reasonable reader could mistake it for a real event or person.
3. Originality and useful publisher content
- No copying, scraping, lightly paraphrased aggregation, copyright infringement or counterfeit promotion.
- Every page needs substantial publisher content. Do not publish placeholders, doorway pages, duplicate variants, misleading downloads or pages created mainly to display ads.
- Headlines, thumbnails and metadata must accurately represent the article; no deceptive clickbait or fabricated urgency.
- Credit licensed material and retain evidence of ownership, licence, permission or applicable legal use for every image, video and substantial extract.
4. Content that must not receive Google ads
Do not monetize content prohibited by Google Publisher Policies, including illegal content; intellectual-property abuse; sexually explicit content; child sexual abuse or exploitation; dangerous or derogatory content; animal cruelty; deceptive practices; malicious or unwanted software; or demonstrably false claims that could undermine participation or trust in elections. When uncertain, classify the article No ads and escalate it for review.
5. Restricted and sensitive subjects
Some lawful journalism may be publishable but receive limited or no advertising. Classify it Restricted subject when it materially involves sexual content, shocking or graphic material, explosives, guns or weapons, tobacco, recreational drugs, alcohol sales or misuse, gambling, prescription drugs, or unapproved supplements. Avoid graphic imagery unless essential to the public interest.
For wars, disasters, deaths, public-health emergencies and other sensitive events, do not exploit suffering, use sensational imagery, encourage profiteering or publish harmful misinformation. Consider No ads even when the journalism itself is legitimate.
6. Health, finance, safety and civic information
- Use authoritative sources and qualified expert context. Do not provide dangerous instructions, guaranteed outcomes or personalized professional advice without appropriate review.
- Do not contradict authoritative information about voting procedures, election eligibility, census participation or official results without exceptionally strong sourcing and precise attribution.
- Clearly disclose uncertainty, conflicts of interest and the limits of available evidence.
7. SEO and promotion
- Write for readers first. The page title, visible H1, Open Graph title, X title and structured-data headline must describe the same story accurately. Each page needs a unique, concise title and useful meta description; length targets are editorial guidance, not a reason to distort meaning.
- Use one natural focus topic plus relevant phrases, entities, category and tags. Keywords support reader discovery and NewsArticle data; the obsolete meta-keywords tag is not a Google ranking signal.
- Publish one HTTPS canonical URL on the HeadlineDecoded domain. Canonical, Open Graph URL, structured data, sitemap and internal links must agree. Search pages, empty sections, placeholders, duplicates and other thin pages must be noindex and excluded from sitemaps.
- Every article needs a descriptive URL slug, verified byline, publication and modification dates, clear source attribution, useful internal links and structured data that exactly matches visible content.
- Lead and sharing images must be owned, licensed or permitted; crawlable; relevant; descriptively named; supplied with meaningful alt text; and preferably at least 1,200 pixels wide for large previews. Clearly label generated or illustrative images.
- Publication must generate complete static metadata, NewsArticle and breadcrumb structured data, Open Graph and X cards, the standard sitemap, Google News sitemap and RSS feed. Resolve fatal SEO audit errors before publication.
- Do not keyword-stuff, create near-duplicate pages, hide text, buy deceptive traffic, send unsolicited promotion or promise rewards for ad views or clicks.
- Never encourage readers to click ads or visually disguise ads as navigation, downloads or editorial content.
8. Mandatory pre-publication checklist
- Originality and meaningful reader value confirmed.
- Facts, claims, quotations and sources checked against the cited evidence during generation.
- Copyright, licence and image provenance confirmed.
- Publisher prohibited/restricted-content review completed.
- AI-assisted material reviewed, corrected and classified by the editorial compliance workflow.
- Ad suitability set to Standard, Restricted subject or No ads.
- SEO title, description, canonical, focus topic, slug, category, image rights, image alt text and social preview checked; generated HTML passed the site SEO audit.
9. The HeadlineDecoded reader and visual standard
Core promise: Understand the news. See the evidence. Remember what matters. HeadlineDecoded primarily serves young readers, especially ages 13–24, while keeping public-interest reporting useful and accessible to everyone.
- Use a youth-first, mobile-first approach: explain how the story affects education, early careers, money, technology, identity, culture, climate, civic life or everyday decisions when the evidence supports that connection.
- Follow evolving conversations on youth-used social platforms as reporting leads. Verify the original post, account, date, context and scale before describing something as a trend; distinguish online reaction from established fact and never manufacture virality.
- Adapt headlines and promotional copy naturally for each social platform without forced slang, clickbait or fabricated urgency. Protect minors and apply heightened care to their identities, images and personal data.
- Use direct, plain language, short sentences where practical and active voice. Explain specialist terms on first use with a concrete example. Preserve nuance: clarity must never remove material context, uncertainty, disagreement or risk.
- Answer the reader’s essential questions: What happened? Why does it matter? How does it work? What evidence supports it? What remains uncertain? What happens next?
- Expand abbreviations on first use whenever practical. If any short form remains in the headline, metadata, visual, quotation or article without a nearby full form, add an Abbreviations and full forms section at the bottom of the article. Include only abbreviations actually used, keep entries alphabetical, and never guess an ambiguous expansion.
- Every future full article must include The story in 60 seconds near the top with four to six concise, verified bullets. The AI Summary must appear immediately below that block, with no other module between them. For a substantial story, also include the useful elements supported by the evidence: a one-sentence decode, “Why it matters”, the deeper explanation, “What we know and do not know”, “What happens next”, and three memorable takeaways.
- Do not pad short updates to fill the complete format. Omit a module when it would repeat information, speculate, trivialise a sensitive event or add no reader value.
Visual explanation requirement
- Assess every major concept for a visual explanation. Add a timeline, process diagram, map, comparison table, verified data chart, cost breakdown, relationship map, before-and-after view or claim-versus-evidence card when it makes the story materially faster to understand or easier to remember.
- In a long or paragraph-dense article, place a few definition or concept diagrams between sections at the exact point where they clarify the surrounding text. Use them to define a term, show a formula, explain a process or compare related ideas. They must not be decorative breaks or repeat the paragraph without adding understanding.
- A useful visual is required for substantial explainers, investigations, data stories and fact checks when verified information supports one. Breaking briefs and simple updates may omit it. Never force a decorative graphic into every section.
- Every visual must teach something, use verified article data, work on mobile, and include a descriptive title, caption, source, methodology or explanatory note where relevant, and meaningful alternative text.
- Never invent data to complete a chart. Clearly distinguish estimates, projections and illustrations. Label AI-generated or reconstructed imagery whenever a reasonable reader could mistake it for documentary evidence.
- Use accessible HTML, CSS, SVG or an optimized static image that remains embedded in the generated public HTML and does not require a database connection.
- For every future substantial article, identify natural visual insertion points during generation and create the required inline-visual package: stable placeholder key, image or deterministic diagram, visual type, descriptive alt text, caption, source or credit, and verified data inputs. Insert visuals with [[visual:key-name]]; never paste unreviewed image HTML into the article body.
- Aim for one meaningful inline visual every two or three major sections when the reporting supports it. Do not add an image after every paragraph, repeat the lead image, or use a visual merely to increase page length.
Trust, memory and retention
- Add a “How we know this” source panel to evidence-heavy reporting. Separate confirmed, supported-but-incomplete, disputed, unverified, misleading and false claims without overstating certainty.
- For an evolving story, show what changed, connect earlier coverage and identify the next verifiable event or date. Do not manufacture urgency merely to encourage return visits.
- Where useful, offer 30-second, three-minute and full-decode reading depths within one canonical article rather than creating thin duplicate pages.
- Optional quizzes, definitions, bookmarks, follows and briefings must help comprehension or establish a voluntary reading habit. Do not use dark patterns, forced registration, misleading next buttons, excessive pop-ups or artificial pagination.
- Recommend related coverage because it adds subject context, not merely because it may increase clicks. Treat completion, saves, follows, source-panel use and genuine return visits as better retention signals than raw page views.
Required generation package
When AI assists a full article, it must return the article with a four-to-six-point The story in 60 seconds block, followed in the public layout by the AI Summary. It should also return, where supported by the reporting brief, a one-sentence decode, why-it-matters text, visual recommendation with verified inputs, visual title/caption/source/alt text, evidence-status notes, three takeaways, glossary terms, related-story ideas, platform-specific social adaptations and complete SEO metadata. It must classify the single primary audience as target_age_group using exactly one value: 13-17, 18-24, 25-34, 35-44, 45+ or all_ages. Flag missing evidence instead of fabricating any component.
10. Site-level AdSense readiness
- Maintain working navigation, substantial original content, clear bylines, About, Contact, Privacy Policy and correction/contact routes.
- Disclose data collection and cookies used by Google products and obtain consent where applicable.
- Keep ads distinguishable from content, avoid accidental-click layouts, do not obscure content with ads, and never place more ads than publisher content.
- Review the AdSense Policy Center and official policies regularly; compliance responsibility remains with the publisher.
Official references
- Google Publisher Policies
- Google Publisher Restrictions
- AdSense Program policies
- Google Search guidance on generative AI content
- Google Article structured data
- Google title-link guidance
- Google snippet and meta-description guidance
- Google canonical URL guidance
- Google sitemap guidance
- Google Images guidance
- Open Graph protocol