Google patented trust. Not figuratively — literally: a patent covered in Harry Clarkson-Bennett's newsletter describes Author Score and Evergreen Score as computable metrics, not guideline suggestions for quality raters. If your copywriter brief still reads "rewrite the top 3 in your own words", it's mathematically obsolete, not just outdated.
A top-3 compilation no longer ranks
In How to Write Great Content for Search (that delivers), Clarkson-Bennett names the diagnosis: AI made compilation free, so only content with genuine Information Gain survives. Three consequences follow, and they apply to every SEO practitioner today.
Behavioral signals decide. Dry copy sends readers pogo-sticking back to the results page — a negative signal to the algorithm. The engaging side of the "4 E" formula holds dwell time and scroll depth, exactly what Google measures.
E-E-A-T is a shield, not a checklist. First-hand experience and expertise protect against demotion more than any technical fix. A first-person account and original research prove the text was written by a human for humans, not by a link-selling site network.
The first traffic spike is a ranking factor. Discover and core search both watch social and newsletter traffic in the first hours after publishing. Distribution is no longer "after publishing" — it's part of publishing.
What that looks like as a formula
Clarkson-Bennett's second piece, Google's Evergreen Article Scoring System Uncovered, breaks the patent down in detail. Five findings that change the brief itself:
| What the patent found | What it means for the brief |
|---|---|
| A hard Commercial Score: aggressive CTAs and pricing in an informational article exclude it from the "in-depth article" index | An informational blog stays informational; the sell lives elsewhere, handled tastefully |
| A one-time link spike reads as manipulation; Evergreen Score grows from steady velocity | Link building is a multi-year process, not a launch-day burst |
| Subfolder Authority: a new article in a strong section gets a starting ranking bonus | Publishing needs a clear hierarchy of sections, not content scattered across the site |
| Author Score is a patented, actually computed factor | Author markup, an on-site profile and external mentions aren't "nice to have" — they're a ranking factor |
| The algorithm tells a deep semantic block apart from a shallow rewrite by text structure | The brief needs complete semantic blocks that cover the topic fully, not a keyword list |
Google confirms: SEO is going multimodal
In March 2025, Google's VP of Ads and Commerce Vidhya Srinivasan described the shift toward visual and multimodal search. For content strategy, that means:
- Search via Google Lens and Circle to Search demands top-quality images with alt attributes and markup — page text is no longer the only entry point into the results.
- AI Overviews capture the most valuable commercial audience — and only cite structured answers: comparison tables, lists, expert content that's easy to lift whole.
- E-commerce SEO has merged with the Merchant Center feed: without full
Productmarkup and every attribute, AI can't match your product to the query. - A "stable top 10" is dissolving under hyper-personalization — the focus shifts to topical authority and brand recognition in independent sources.
- Video is becoming a mandatory entry point: YouTube is integrating into the buying journey alongside text.
In practice: a brief for a real page
Let's work an actual published page instead of an abstraction — Thanksgiving Music Classroom Activities on anastasiya.studio, my wife's store. Yes, the same site where we built the author entity and earned AI citations — identity signals and the depth of one specific article solve different problems, and this page shows what's still left to fix at the content level.
Weaknesses against all five patent findings:
- Thin content. Five activities are described at surface level, with no differentiation scenarios for different grades — exactly what the patent flags as a rewrite signature rather than depth.
- Generic tone. "Wonderful way to teach", "ideal activity" — vague praise with no authorial voice and no single real classroom moment.
- Zero E-E-A-T. No author credentials, no teacher testimonial, no proof the activities were actually run in a classroom.
- Obvious keyword targeting. "Thanksgiving Music Worksheet" repeats mechanically across tags and product titles — easy to read as optimized for a robot, not a person.
- Commercial Score risk. Five product bundles and a "get one free" block sit right under the text — the information/sell balance needs to become a clean, native integration.
What follows isn't a shortened example — it's the actual rewrite brief, exactly as I'd hand it to a copywriter for this specific page.
Rewrite brief: Thanksgiving Music Classroom Activities
Page goal: an informational article for music teachers that also leads, natively, into five product bundles — with no pressure on the reader inside the body text.
Target reader: a K-5 music teacher looking for ready-made Thanksgiving lesson ideas; secondary audience — Pre-K teachers.
1. Structure and length. Keep the H1. Before the list of activities, add a 60-80 word intro paragraph: why these five, and what ties them together (there is currently no intro paragraph carrying the author's own point of view at all). Each of the five activities gets a 150-200 word block (currently 60-80): description, step-by-step execution, expected outcome.
2. First-hand experience is mandatory in every block, not one general chunk up top. Each activity needs its own real classroom moment: how it actually went, what the kids said, what had to change mid-lesson. Not "a wonderful way to teach" — instead: "when we played freeze dance over the song's quiet part, half the class froze at the wrong beat, so we added a counted-out warning for the first three rounds." Without this, the block stays exactly the "generic tone" the patent distinguishes from real experience.
3. Age differentiation. Each activity gets its own "How to adapt it" line: one sentence for K-2, one for grades 3-5. The patent specifically reads this kind of detail as a depth signal, not a shallow rewrite.
4. Keywords — only from real queries. Before writing: pull every query this page already ranks for from Google Search Console (Performance → Pages → this page → Queries). Weave the actual phrasing of those questions into the text — not as a keyword list tacked onto a paragraph, but dissolved into the explanation. Check specifically whether "Thanksgiving Music Worksheet" repeats mechanically three times in a row — it currently does, and it's the most obvious tell of optimizing for a robot.
5. Meta description — using the method from "How to Write a Meta Description That Google Will Use": one question the whole page answers → a short answer → reworded as a standalone description → weave in the GSC queries → fit it into 157 characters with the substance in the first 120 → one imperative verb.
6. Image alt text. Each activity gets a photo or illustration with a substantive alt (not "thanksgiving music activity" — instead "kids doing freeze dance during the slow part of a song in music class") — in line with Google's shift toward visual search covered earlier in this same article.
7. E-E-A-T. Byline the article with the author's name linking to an About page with credentials. If you can get one, add a short testimonial from another teacher who used the material, naming them and their school (or at least their state).
8. Commercial balance. No product names as section headings, no CTAs inside the instructional paragraphs. The five product bundles live in one clearly separated block after the text, with nothing woven into the narrative.
9. Ending. Instead of a buy prompt, one practical takeaway: what the teacher can do this Monday morning, using one idea from the article without buying the bundle.
The results of this brief — Google Search Console screenshots before and after, ranking changes, and an AI Overviews citation if it lands — go in a separate piece I'll fill in as the rewrite ships: "Rewriting the Thanksgiving Music Activities Page: What Happened."
The takeaway
Three sources — a practitioner's newsletter, a Google patent, and a VP's own statement — converge on one point: content is now scored as a system of identity, behavior and structure signals, not as text optimized for keywords. Compiling someone else's results page is detected by the algorithm and fails. First-hand experience, steady distribution and author markup are counted — and that is exactly the part no model can generate without you.
0 comments
No comments yet — be the first to share your thoughts.
Sign in to post your comment instantly:
…or comment as a guest — guest comments appear after moderation.