Jonathon's AI Wiki

Content Effort — Google's Quality Signal and the contentEffort Score

6 min read

Source: raw/newsletter-zyppy-signal-5fcdd4b44a.md

Author: Cyrus Shepard (Zyppy Signal) | URL: https://signal.zyppy.com/p/content-effort | Published: 2026-08-13

Shepard — a former Google Quality Rater — documents effort as an explicit scored dimension in Google’s Quality Rater Guidelines, alongside originality, talent/skill, and accuracy. The Google API leak surfaced two matching fields, contentEffort and ugcDiscussionEffortScore, and the crucial reframe is that effort is not how much work you did — it is the on-page evidence of useful work. That distinction is what makes the signal actionable rather than moralistic.

Key Takeaways

  • Effort is a named, repeatedly-scored dimension. Raters evaluate every page on effort alongside originality, talent/skill, and accuracy. The word appears over 100 times in the Quality Rater Guidelines, and Google nearly always discusses it in the same breath as originality and talent/skill — the three overlap heavily.
  • Google cannot measure your effort and does not try to. It estimates it from visible evidence, against strict standards, drawing on thousands of human-rated documents, machine learning at scale, and now LLMs.
  • Two leaked fields sit in compressedQualitySignals — page-specific signals the docs say “can be used in preliminary scoring,” alongside lowQuality and siteAuthority. contentEffort is a floating-point number (an actual numeric score) and is documented as LLM-produced, potentially separate from Quality Rater scoring. ugcDiscussionEffortScore covers user-generated pages.
  • The honest caveat is stated up front: we do not know how, or even whether, Google uses these metrics. What the documentation, public statements, and leak jointly establish is that Google spends substantial effort evaluating effort.
  • Effort is explicitly NOT: word count · time spent creating the page · human-written being automatically better than AI. All three are the intuitive proxies, and all three are wrong.
  • Automation is not penalized — undifferentiated automation is. The test is value-add that automation cannot easily replicate: proprietary database content, human evaluation and perspective, quality UGC, unique media, authentic reviews. Low-effort: 5,000 templated location pages with the city name swapped. High-effort: automated recommendation pages built from real reviews in an app you own plus proprietary real-time ticket data.
  • First-party information is the load-bearing concept. Content reproducible by anyone with AI, Google, and Wikipedia is low-effort by construction. Value has to originate with the page’s creator — original facts from your own experience, plus genuine perspective. Shepard calls Google’s move toward “non-commodity” content one of the most significant algorithmic shifts in years.
  • Media counts as content. Stock or borrowed images and video — including lightly altered ones — read as low-effort. Original photography and video do not guarantee rankings but do demonstrate the work.
  • UGC is scored on participation depth, not authorship. Contributions from multiple people accumulate into real total human effort; a thread with one “Yep, me” reply does not.

The seven effort signals

Shepard classifies Google’s guidance into seven, each with a paired example:

#SignalLow-effortHigh-effort
1Curation and editingChicken recipes scraped from other sitesChicken recipes personally tested, organized by popularity and taste
2Automation vs added value5,000 templated location pages, city name swappedAutomated recommendations built on owned user reviews + proprietary ticket data
3Recycled info vs original facts”10 Best Protein Powders” describing 10 popular powders”10 Best Protein Powders We’ve Tested” — methods, results, opinions, detailed photos
3bRecycled vs original mediaStock or lightly-altered borrowed imagesDetailed original photography and video
4Filler vs prominent helpful content”How to boil an egg?” opening with 600 words on egg historyKey facts up front, then egg size, altitude, yolk firmness
5Shallow vs quality discussionForum post with a single “Yep, me” replyBoot reviews with wear experience and reasoned recommendations

Google’s own definition of filler: “low-effort content that adds little value and doesn’t directly support the purpose of the page. Filler can artificially inflate content, creating a page that appears rich but lacks content website visitors find valuable.” The classic SEO sin — keyword-rich preamble for topical relevance — is filler by this definition, and it is worse when placed at the top of the page to drive ad views or time-on-site.

Try It — the two-minute pre-publish audit

Shepard’s five checks, to run before creating or publishing:

  1. Is the content edited and curated thoughtfully? Organized around the user’s goal, most valuable content easy to find.
  2. Did you add something that doesn’t exist elsewhere? Original data, facts, observations, opinions, perspectives.
  3. Does the content show evidence of work? Stated methodology, high-quality photos, screenshots, interviews, observations.
  4. Does it go beyond common facts? If someone could reproduce it by researching the internet, it is low-effort. Go to details and edge cases.
  5. Does it provide deep discussion? Not every page needs it (a stock price does not), but where it applies, considered evidence plus a rendered judgment is what elevates a page.

The operational reframe worth internalizing: you are not being asked to work harder, you are being asked to leave proof on the page. A genuinely researched article with no methodology section, no original media, and no stated opinion can score as low-effort — and a modest page with one original test and a real photograph can score higher.

Open Questions

  • Is contentEffort actually used in ranking? The source is explicit that this is unknown. It sits in a structure documented for “preliminary scoring,” which is suggestive, not confirmation.
  • What is the score’s range and distribution? Documented as a float; no scale, threshold, or calibration is public.
  • Is the LLM-produced contentEffort consistent with human rater judgments? The docs indicate it may be separate from rater scoring, which raises the question of whether the two agree.
  • How does effort interact with AI-search citation? The wiki’s citation cluster measures what gets cited by LLMs, not what Google’s effort scorer rewards. Whether high-effort pages are cited more is untested here.
  • The audit scoring system (0–100, 50+ evidence examples, per-vertical playbooks) is paywalled in Zyppy Pro Templates and is not captured in this article. The companion post raw/newsletter-zyppy-signal-084898bc4d.md is a pointer only.
  • Leak-derived field names are not documentation. contentEffort and ugcDiscussionEffortScore come from the Google API content warehouse leak, not from Google’s published guidance.

Graph View

Backlinks