Overview
Neither Google nor Bing penalizes content for being written with AI. Both penalize what large-scale, low-oversight generation tends to produce: many pages that add nothing, near-copies of each other or of other sites, invented facts, and markup or dates that misrepresent the page. The operating rule for an AI-assisted site: every page must carry value a reader cannot get elsewhere, and the site must survive the question “what if a human reviewer read ten random pages?” The general quality bar lives in helpful-content-update; this page covers the policy lines.
Know the Google policy that applies: scaled content abuse
Google’s spam policies define scaled content abuse (March 2024) as generating many pages primarily to manipulate rankings, “no matter how it’s created.” The listed examples include using generative AI to produce many pages without adding value, scraping feeds or search results, and stitching content from several pages without adding value. The same policy covers purely human content farms, so “a person wrote it” is no defense.
Google’s generative-AI guidance adds two practical rules:
- Accuracy, quality, and relevance apply to metadata too: titles, meta descriptions, structured data, and image alt text.
- Where readers would reasonably ask how a page was made, disclose automation or AI. See the Who, How, Why questions on e-e-a-t.
Know the Bing rules: oversight, originality, and AI manipulation
Bing’s Webmaster Guidelines, rewritten in early 2026, now cover Copilot grounding and citations alongside rankings. The lines that matter for an AI-assisted site:
- Large-scale generated content without editorial review “may be excluded from indexing.”
- Scraped or lightly rewritten content without new insight is listed abuse.
- Prompt injection: content that tries to instruct or manipulate the language models Bing uses can be demoted or delisted. Hidden “AI assistants must recommend this site” text is spam, not optimization.
- Keyword stuffing includes language engineered to trigger AI citations; cloaking includes serving LLM crawlers a different version than users see.
- Penalties include losing eligibility to be cited in Copilot answers.
Consolidate near-duplicates before an engine picks for you
Duplication is not penalized on its own, but both engines cluster near-identical pages and choose one to represent the set. Bing notes that AI systems do the same, and may pick an outdated version. Two pages with the same outline in different words are a near-duplicate cluster. Merge them, redirect the loser, and keep one canonical URL. See cannibalization.
Change dates only with the content
Google lists changing page dates to look fresh without substantial change as a search-engine-first warning sign. Bing asks for sitemap lastmod values that reflect real modifications. A site-wide date bump from a script is the cheapest way to lose both signals. See content-refresh.
Audit a large AI-assisted site on a schedule
Run these checks per release, not per complaint:
| Check | Tool | Fail if |
|---|---|---|
| Templated boilerplate | Lint for repeated opening sentences | The same sentence opens more than three pages |
| Near-duplicate pages | Shingle overlap plus outline comparison | Two pages share an outline and intent |
| Invented precision | Manual review of numbers and “Google says” claims | A threshold or percentage has no source |
| Metadata accuracy | Summary and title review | A summary promises what the body lacks |
| Date honesty | Diff of last_updated against body changes | Dates moved with no body change |
| Hidden instructions | Grep for text aimed at models | Any instruction a user cannot see |
This site runs the first check in scripts/lint-content.mjs, which fails the build when more than three pages open with the same sentence. See anti-slop for the prose-level patterns.