Guide · Helpful content & E-E-A-T

Google's Helpful Content Guidance: A Working Rubric for Content Teams

Google's people-first questions turned into a rubric a content team can score pages with: eight dimensions, what Strong looks like on each, the warning signs Google lists, a calibration routine and a worked example.

Ulrich Svarrer

November 15, 20269 min read

Google’s helpful content guidance is a page of self-assessment questions about whether your content is made for people or for search engines. This guide turns those questions into a rubric a content team can score pages with: eight dimensions, five levels, what Strong looks like on each, the warning signs that should stop a page from publishing, a calibration routine so two editors score the same page the same way, and a worked example. It is the rubric we hold our own guides to, and the same eight dimensions Morrison’s content analyzer scores.

The guidance is Google’s Creating helpful, reliable, people-first content. Read it once in full; it is short. What it lacks is a way to apply it consistently across hundreds of pages and several writers, and that is what the rest of this guide is for.

What the guidance is now, and what it is not

Much of what you will read about “helpful content” is out of date. The facts, from Google’s own documentation:

There is no separate helpful content system any more. Google launched it in 2022 as the “helpful content update”. In March 2024 it became part of the core ranking systems, and Google’s ranking systems guide now lists it under retired systems. Google said at the time that there is no longer one signal or system used to assess helpfulness.
The questions are for you, not a checklist of ranking factors. Google offers them to help you evaluate your own content, and points to them when sites ask what to do after a core update.
E-E-A-T is not a ranking factor in itself. Google says its systems use a mix of signals that align with it, that trust matters most, and that it weighs more heavily on topics that can affect health, finances, safety or society. See E-E-A-T for content teams.
Quality raters do not rank pages. Google says rater data is not used directly in its ranking algorithms; raters help it check whether its systems work. The rater guidelines, last updated 11 September 2025, are still the most detailed description of what Google means by quality.

So the guidance is best used as a quality standard: what a page has to be to deserve the traffic you want it to earn. Used that way, it is also the most practical one available, because it is written by the company whose results you are trying to appear in.

What the guidance asks

The page groups its questions into a few sets. In brief, paraphrased:

Group	What it asks
Content and quality	Original information or analysis; a substantial, complete treatment; insight beyond the obvious; real added value when drawing on other sources; an accurate, non-exaggerated title; whether readers would bookmark or recommend it; whether it is well produced, not mass-produced
Expertise	Clear sourcing and evidence of expertise; author and site background; whether an expert wrote or reviewed it; no easily verified factual errors
Page experience	A good experience overall, across many aspects rather than one or two
People-first	An audience that would value it if they came directly; first-hand experience; a clear main purpose; readers learn enough to reach their goal and leave satisfied
Search-engine-first warning signs	Made mainly for search traffic; many unrelated topics; heavy automation; summarizing others without adding value; chasing trends; readers have to search again; writing to a word count; changing dates or content just to seem fresh
Who, How and Why	Who made it, and is that clear; how it was made, including any automation; why it exists, which Google calls the most important question

The page also defines main content as everything that directly serves the page’s purpose, such as text, tools and headings, and says that fabricated creator profiles, such as invented names, AI-generated headshots or false credentials, are deceptive. Both are worth building into any rubric: the first decides what you score, the second is a hard stop.

From questions to a rubric: eight dimensions

Questions are hard to score. Dimensions are easier. We group Google’s questions into eight dimensions, each with a description of what Strong looks like and the evidence a reviewer should look for. The names match the dimensions in Morrison’s content analyzer, so scores from people and from the analyzer can be compared.

Dimension	Google’s questions it covers	Strong looks like	Evidence to check
Originality	Original information, insight beyond the obvious, value beyond the sources	Something the top results do not have: own data, a worked example, a framework, a tested method	Compare with the top five results: what here is not there?
Effort	Substantial, complete; well produced, not rushed or mass-produced	Concrete steps, examples, tables or figures where they save the reader time	Could a competent person produce this in an hour from the top results?
Helpfulness	Reader learns enough to reach their goal; no need to search again	Answers the task end to end; the first paragraph says what the page answers	Read it as the searcher: what would you still need?
Skill	Evidence of expertise, written or reviewed by an expert, first-hand experience	A named author or reviewer with relevant background; specifics only practice gives	Byline, author page, and whether the details are specific or generic
Accuracy	No easily verified factual errors; accurate title	Every claim correct today, important ones sourced to primary documents	Spot-check five claims against their primary sources
Trust	Clear sourcing, background about the site, honest presentation	Sources listed, commercial interests clear, dates honest, no made-up results	Sources, disclosures, dates, about and contact information
People-first	An audience that would come directly; clear purpose; satisfying	Written for a defined reader doing a defined task; no padding	Would you send this page to a customer who asked the question?
Spam	The warning signs, and Google’s spam policies	No warning signs: not scaled, not stuffed, not a near-duplicate, honest dates	The warning-sign gate below

Five levels

Score each dimension on five levels. Short definitions keep scorers aligned:

Level	Meaning
Weak	Fails the dimension; would embarrass the brand if a customer read it
Limited	Present but thin; the reader would notice what is missing
Fair	Adequate, and no better than what already ranks
Strong	Clearly better than average on this dimension, with evidence on the page
Excellent	The best available answer on this dimension; others would cite it

The bar we use for our own guides is Strong or better on all eight. A page that is Excellent on Accuracy and Fair on Originality is a correct restatement of what already ranks, and Fair is where most content sits.

The warning signs, as a gate

Google’s search-engine-first warning signs work best as a gate before scoring: if one applies, fix it first, because no amount of quality elsewhere outweighs it. Our rule is that any open warning sign caps the page at Limited on People-first and Spam. Here is what each looks like inside a content team:

Warning sign	How it shows up in a content team
Made mainly to attract search visits	A keyword list drives the calendar, and nobody can name the reader
Lots of topics hoping some perform	Pages far outside what the company knows, because the volume looked good
Extensive automation across many topics	AI drafts published with light edits and no subject review
Mainly summarizing others	Briefs that say “cover what the top five cover”
Writing to a word count	A target length in the brief that has nothing to do with the task
Readers need to search again	The page explains the concept but leaves out the steps, figures or template the searcher came for
Promising an answer that does not exist	“Release date” or “price” pages for things not announced
Changing dates to seem fresh	A CMS or habit that re-dates pages without real changes; see what counts as a real content update

Several of these overlap with Google’s spam policies, which go further. Content made at scale mainly to manipulate rankings is scaled content abuse, however it is produced. See Google’s spam policies for content teams.

Who, How and Why, in practice

Who. A byline where readers would expect one, linked to an author page that explains the person’s relevant background. Never an invented author or credential: Google’s guidance now names fabricated creator profiles as deception. If a specialist reviewed the page, say so and name them.
How. Where readers would reasonably wonder how a page was made, tell them: how a product was tested, where data came from, whether AI was used substantially. Google’s guidance on generative AI content asks you to fact-check AI output, including titles, descriptions and structured data, and to consider disclosure.
Why. The question Google calls most important. Write the answer in the brief: “This page exists to help [reader] do [task].” If the honest answer is “to rank for this keyword”, the page has a People-first problem before it is written.

Scoring in practice

Calibrate before you score at scale

Two editors with the same rubric will still disagree at first. Fix that before scoring hundreds of pages:

Pick ten pages that span your range: two you consider your best, two your weakest, six in between.
Have two or three people score all ten independently, with one line of evidence per score.
Compare. Any dimension where scores differ by more than one level gets discussed until you agree what the evidence means.
Write down one anchor example per level for the dimensions that caused most disagreement. These become the reference for future scorers.
Repeat with five new pages each quarter, or when someone new joins.

Where to use the rubric

The rubric across a page's life

Brief

Set the Originality and Skill plan: what is new, who brings the expertise

Draft review

Score all eight; anything below Strong goes back with the evidence

Pre-publish gate

No open warning signs; Accuracy spot-check done

Audit

Score the library; lowest scores on high-value pages go first

Worked example

An illustrative example, not client data: a guide to holiday pay rules for employers, ranking around position 7 for its main query.

Dimension	Score	Evidence	Fix
Originality	Fair	Restates the rules the top three results explain	Add a worked calculation for a part-time employee
Effort	Fair	Text only; no table of rates or example	A rates table and the worked example
Helpfulness	Fair	Explains the rule, not how to apply it to a leaving employee	A section on leaving employees
Skill	Fair	Byline is “Marketing team”	Name the payroll specialist who reviewed it
Accuracy	Strong	Figures match the official source	Link the official source next to the figures
Trust	Strong	Sources listed; product mentioned once, clearly	None
People-first	Strong	Written for employers with a clear task	None
Spam	Excellent	No warning signs	None

The pattern is typical: accurate and honest, but no better than what already ranks. The four fixes are a day’s work with the payroll specialist, and each one adds something the top results lack, rather than more of what they already have.

Morrison reads the stored helpful-content scan for one guide, scores the top results with the same analyzer and names the weakest dimensions and the fixes.

Common misreadings

“Longer is more helpful.” Google says it has no preferred word count. Completeness is about the task, not length.
“Update the date to look fresh.” A named warning sign. Update the content; the date follows.
“An author bio fixes E-E-A-T.” A bio is evidence of Skill only if the page shows the expertise too.
“AI content is penalized.” Google judges content by its quality and purpose, not how it was produced; mass-producing pages mainly to rank is the problem.
“It is a sitewide classifier we have to escape.” That system no longer exists as a separate system; improve the pages that matter most and keep going.

Where Morrison fits

Morrison’s content analyzer scores every page in your site index on these eight dimensions after each crawl, with the same five levels. Each page’s detail view shows its scores and which of Google’s questions it meets, partly meets or misses. The URL grid sorts by the overall score and can show a column for every dimension; export it to CSV to sort a whole section by Originality or Trust. That turns the audit step from a sample into the whole site. Chat can rank pages by the overall score, read their scans and explain a score by comparing the page with live results, and it can score a competitor’s page the same way. The human parts stay human: calibrating what Strong means for your audience, and adding the original work and expertise that move a score. See the content analyzer and E-E-A-T assessment.

Key takeaways

It is guidance, not a system to beat: the helpful content system is now part of core ranking, with no single signal.
Turn questions into a rubric: eight dimensions, five levels, evidence for every score.
Gate on the warning signs before you score anything else.
Calibrate, or two editors will score the same page two levels apart.
Aim for Strong on all eight; Fair is what already ranks.

Sources and further reading