Google publishes two documents that set out, in plain language, what it treats as quality content: the Search Quality Rater Guidelines, a 182-page manual written for its human evaluators, and the "Creating helpful, reliable, people-first content" page in its Search Central documentation. Neither hands you a formula, but both describe concrete criteria any organisation can apply to its own pages.
What follows works through those criteria in their current form: the rater guidelines dated 11 September 2025, the E-E-A-T model, the spam policies, and the core updates shipped across 2025 and 2026 — along with a few persistent misreadings, starting with the one that equates quality with length.
One framing matters throughout: these documents do not describe how the ranking systems work, only what those systems are trying to reward.
What the rater guidelines are, and what they are not
The Search Quality Rater Guidelines (11 September 2025 version, 182 pages) are written for external evaluators who rate search results. The document says so itself: no individual rating moves a page up or down; the ratings measure whether the algorithms are doing their job.
The scale is considerable. On its Rigorous testing page, Google reports running 719,326 search quality tests and 124,942 side-by-side experiments in 2023, leading to 4,781 changes launched that year (Google's own figures, worldwide, 2023). The guidelines are the yardstick for those adjustments, not a rulebook applied page by page.
The manual rests on two separate ratings:
- Page Quality: how well a page achieves its purpose, on a scale running from Lowest to Highest.
- Needs Met: how well a result serves the need expressed by the query, from Fails to Meet to Fully Meets.
The appendix change log shows the recent direction of travel: the Lowest and Low sections were aligned with the spam policies in January 2025, then the YMYL definitions — "your money or your life" — updated in September 2025.
Four criteria for judging the main content
Section 3.2 of the guidelines defines main content quality through four ideas, each describing something a human reader can perceive.
| Criterion | What Google asks raters to assess | What it means for a business page |
|---|---|---|
| Effort | Did a person actively work to produce content that serves the visitor? | A page built from your own cases, edited and structured, not a generic text. |
| Originality | Does the content offer something that does not exist elsewhere? Who is the source? | Your pricing logic, your lead times, your own photographs, what you learned on the job. |
| Talent or skill | Is the content made with enough competence to satisfy the visitor? | A clear explanation, no unnecessary jargon, readable visuals. |
| Accuracy | Is the content factually correct and, on YMYL topics, consistent with expert consensus? | Dated and sourced figures, current regulatory references, claims you could defend. |
One principle governs all four: the page must achieve its purpose. A product page, a blog article and a contact page are not held to the same expectations, and a small independent site is not judged against the resources of a large one.
E-E-A-T: trust at the centre
E-E-A-T stands for experience, expertise, authoritativeness and trust. The guidelines are explicit that trust is the most important member of the family. The other three exist to support it.
- Experience: first-hand practice with the subject. Google's example is a product review written by someone who used the product.
- Expertise: the knowledge the topic requires — electrical advice from an electrician rather than an enthusiast.
- Authoritativeness: is the site or author recognised as a reference on this subject?
- Trust: is the page accurate, honest, safe and reliable? For a shop, that extends to secure payment and reachable customer service.
Two clarifications prevent the usual misreading. First, Google states that E-E-A-T is not a ranking factor in itself: its systems use a mix of signals to identify content that shows those characteristics. Second, the level required depends on the subject — on YMYL topics the standard is higher, and the tolerance for a thin author biography lower.
Who, how and why: Google's three questions
The Creating helpful, reliable, people-first content page turns E-E-A-T into three practical questions:
- Who created the content? A byline and short biography wherever a reader would expect one.
- How was it created? Method, evidence, and — where AI played a significant role — a note where a reader might reasonably wonder.
- Why was it created? The expected answer is: to help the people visiting the site. If the primary purpose is to attract search traffic, Google is direct that this is not what its systems set out to reward.
What pushes a page into Lowest or Low
Since January 2025, the bottom rungs of the rating scale draw directly on Google's spam policies. Three practices are named in the manual:
- Scaled content abuse: producing many pages with little effort and no added value for visitors, "regardless of how it's created". The guidelines explicitly mention automated tools, generative or otherwise, as a cheap way to multiply pages.
- Expired domain abuse: buying a domain to exploit its history and publishing unrelated content on it.
- Site reputation abuse: hosting third-party content to trade on the ranking signals of an established site.
Alongside those sit more ordinary failings: copied or lightly reworded content, filler that delays the useful information, too little about the company or the author, keyword stuffing. Generative AI is not banned. Google presents it as a useful tool that can be misused, and judges the output rather than the method. The risk is not in using a model, but in publishing at a volume nobody has read.
Where pages have thinned out this way, the repair work is editorial rather than technical: reworking the content itself moves more than another round of tag adjustments.
Length is not a criterion
A stubborn belief holds that longer texts rank better. Google contradicts it head-on in its self-assessment list, asking whether you are writing to a particular word count because you heard Google prefers one — which, it adds, it does not. What counts is that the reader does not have to go and search again somewhere else.
Core updates in 2025 and 2026: what the calendar shows
Since March 2024, the former "helpful content system" has been part of Google's core ranking systems, so the criteria above are re-weighted at every core update. The Google Search Status Dashboard (consulted in September 2026) lists the following rollouts.
| Update | Rollout start | Approximate duration |
|---|---|---|
| March 2025 core update | 13 March 2025 | 14 days |
| June 2025 core update | 30 June 2025 | 17 days |
| August 2025 spam update | 26 August 2025 | 27 days |
| December 2025 core update | 11 December 2025 | 18 days |
| March 2026 spam update | 24 March 2026 | under one day |
| March 2026 core update | 27 March 2026 | 12 days |
| May 2026 core update | 21 May 2026 | 12 days |
| June 2026 spam update | 24 June 2026 | 2 days |
| August 2026 spam update | 18 August 2026 | 3 days |
If your traffic falls after one of them, Google's documentation advises waiting at least a week after the rollout finishes, comparing the affected pages and queries in Search Console, then going back through the self-assessment questions. Improvements can take months to be reflected, and rushed corrections are discouraged. A diagnosis starts from the fundamentals of search visibility — indexing, structure, and whether the content differs from what already ranks. If the drop looks structural, scope the audit before reacting page by page.
The same criteria apply to AI Overviews and AI Mode. Google states that no extra requirement and no dedicated markup are involved. The page has to be indexed, eligible to appear in search, and useful. No separate bar to clear also means no separate shortcut.
A self-assessment before you publish
Google offers around twenty self-assessment questions. These decide the outcome most often for a smaller organisation:
- Does the page contribute original information, analysis or data, or restate what other sites already say?
- Does the title describe the content accurately, without a promise the text does not keep?
- Is the author identified, with enough for a reader to judge their competence?
- Are sources cited, and are the facts checkable?
- Does the reader leave with what they came for, or search again?
- If AI was used significantly, is that stated where a reader would expect it?
- Are you publishing this because your audience needs it, or because the topic is popular?
Google adds one piece of advice that is rarely followed: have the content read by people you trust with no connection to the site. They notice the filler you stopped seeing weeks ago.
Keep the limits in view. These criteria describe what the algorithms aim at, not how they work. Following them guarantees nothing; ignoring them accumulates reasons to slip.
Common questions
Are the Search Quality Rater Guidelines a ranking factor?
No. It is written for human evaluators who rate results to measure how well Google's algorithms perform, and no individual rating moves a page up or down. The manual stays useful because it describes what Google is trying to reward.
What does E-E-A-T mean, and which part matters most?
E-E-A-T covers the experience, expertise, authoritativeness and trustworthiness of a piece of content and its author. Google states that trust is the central criterion and the other three support it. The expected level is higher on topics touching health, money or safety.
Does Google penalise content generated by AI?
No, AI is not prohibited in itself. Google judges the result: content that is useful, accurate and original can be AI-assisted. Producing pages in bulk with no added value, with or without AI, falls under scaled content abuse, which is a spam practice.
Does a long article rank better than a short one?
Google says explicitly that it does not prefer any particular word count. The right length is the one that lets the reader get their answer without searching elsewhere, and padding is listed among the characteristics of low quality pages.
What should I do if my traffic drops after a core update?
Google advises waiting at least a week after the rollout ends, then comparing affected pages and queries in Search Console and reassessing that content against the self-assessment questions. Improvements can take months to be reflected, so avoid rushed rewrites.
Is special markup needed to appear in Google's AI Overviews?
No. Google states that no additional requirement and no dedicated file are involved. The page needs to be indexed, eligible to appear in search, and useful — the same criteria as for ordinary results.