There’s a word buried in Google’s Search Quality Rater Guidelines that appears more than a hundred times, and most people who read it come away with the wrong idea.

The word is effort.

Cyrus Shepard, who has actually worked as a Google Quality Rater, published an analysis of how the guidelines treat it. The finding that matters is in his subheading, and it’s worth stating plainly before anything else.

Effort is not how much work you put in. It’s the evidence of useful work that ends up visible on the finished page.

Those are very different things, and the gap between them explains a lot of frustrating SEO outcomes. You can spend three weeks on an article and produce something Google reasonably scores as low effort. You can spend an afternoon and produce something scored as high effort. What separates them isn’t the hours. It’s whether anything on the page could only have come from you.

Let me unpack how that works, because it’s one of the more useful lenses I’ve found for deciding what’s actually worth writing.

What Google Actually Says

Raters are asked to consider the extent to which a human being actively worked to produce content that satisfies the reader. The guidelines allow for that work taking different forms. Someone translating a poem by hand counts. So does building the functionality or systems behind a page, such as creating a genuine machine translation service.

What doesn’t count is running thousands of pages of freely available content through translation software automatically, with no oversight or manual curation, and publishing the result.

Notice what distinguishes those cases. It isn’t automation. The machine translation service is automated and still counts. What’s missing in the bad example is human judgment applied anywhere in the process.

Effort also rarely appears alone in the guidelines. It sits alongside originality and talent or skill, and the three overlap heavily. Raters assess them together when judging the main content of a page.

The Part From the Leak

When Google’s internal API documentation leaked, two attributes turned up that relate directly to this.

One is contentEffort, described as an effort estimation for article pages produced by a language model. The other is ugcDiscussionEffortScore, relating to user-generated content pages.

Both sit inside a structure of page-level quality signals that the documentation indicates can be used in preliminary scoring. The same structure holds other attributes with names like lowQuality and siteAuthority. Notably, contentEffort is defined as a floating-point number, which suggests an actual numeric score rather than a simple flag.

I want to be careful here, and Shepard is careful about it too. Leaked attribute names tell us what exists in the documentation. They don’t tell us how any of it is weighted, whether it’s currently active, or how it interacts with everything else. Anyone presenting this as a confirmed ranking formula is overselling it.

What it does suggest is that effort isn’t only a human rater concept. Google appears to have thought about estimating it computationally, at scale. Combined with the guidelines and Google’s public statements, the direction is consistent enough to plan around even without the mechanics.

Three Things Effort Is Not

This is where most people go wrong, and each of these is worth stating separately.

It is not word count. Adding another eight hundred words of context to a page doesn’t raise effort. It often lowers it, for reasons I’ll get to under filler.

It is not time spent. Google has no visibility into your hours and doesn’t try to guess them. A page that took a month and a page that took an hour are assessed identically, on what’s actually on them.

It is not human-written versus AI-written. This one matters given how much anxiety sits around it. The guidelines don’t reward text for being typed by a person. They reward evidence of human judgment and first-hand knowledge in the result. AI involvement isn’t the disqualifier. Absence of anything only you could contribute is.

What Counts as Evidence

Shepard groups the guidance into a set of recognisable signals. Here they are with the examples, because the examples do more work than the categories.

Signal	Low effort	High effort
Curation	Chicken recipes scraped from other sites	Chicken recipes personally tested, ordered by taste
Automation	5,000 location pages, one template, city names swapped	Auto-built pages using owned review data and proprietary ticket data
Original facts	“10 Best Protein Powders” with no testing behind it	“10 Best Protein Powders We’ve Tested” with method, results, photos
Media	Stock or borrowed images, lightly altered	Your own photography and video of real work
Filler	600 words on egg history before the boiling method	Method first, then altitude, size and yolk firmness detail
Discussion	A thread with one reply saying “yes, me too”	Boots reviewed from real use, with a reasoned recommendation

Curation and editing

A low-effort page collects information with little human involvement. A high-effort page shows someone decided what mattered and organised it usefully. Same subject, same format. The difference is that a person made choices and those choices are visible.

Automation with genuine added value

Automation isn’t the problem. Automation producing content that adds nothing over what already exists is the problem. Both examples in the table are automated at scale. The second one has inputs nobody else can replicate, which is the entire distinction.

This is where things like your own product catalogue, your own review data, your own images and your own evaluations become genuinely valuable rather than just nice to have.

Original facts rather than recycled ones

Low-effort content restates information available elsewhere. It’s the material anyone could assemble with a search engine and a language model. High-effort content contains facts you obtained yourself, through doing the thing, plus perspectives that come from having done it.

Media counts as content

This one gets overlooked. Images and video are assessed the same way. Stock photography and images borrowed from elsewhere, including lightly modified ones, read as low effort. Original photography and video read as high effort, because they’re proof you were there.

Good photographs don’t guarantee rankings. But a page with your own images of your own work is demonstrating something that a page with a stock photo cannot.

Prominent useful content rather than filler

Google’s guidelines define filler as low-effort material that adds little and doesn’t support what the page is for. The specific concern is content that inflates a page so it looks substantial while lacking anything a visitor would value.

Every SEO has written the long-preamble version at some point. It was a relevance tactic. It now works against you, and it’s worth going back through older pages with this in mind.

Real discussion rather than shallow discussion

Google appears to want more than a recitation of facts. It wants evidence that someone weighed things and reached a judgment.

For user-generated pages the guidelines take an interesting position: the depth of conversation and the level of participation both contribute, and contributions from multiple people accumulate into meaningful total effort.

Not every page needs this. A page listing current share prices doesn’t need a perspective. But for anything involving a decision, the judgment is a large part of the value.

The Two Questions

Strip all of that down and you get a test you can run in about a minute, before you write anything.

What on this page could only have come from me? What visible evidence shows I actually did the work?

If you can’t answer the first one, you’re producing something reproducible by anyone with a search engine, and you’re competing on nothing. If you can’t answer the second, you may have done the work but the page isn’t showing it, which for Google’s purposes is much the same thing.

The second question is the one people neglect. Plenty of businesses have genuine expertise and publish pages that display none of it. The knowledge exists in the owner’s head and never makes it onto the screen.

What This Looks Like for a Real Business

Abstract advice is easy. Here’s how I’d translate it.

A lawyer writing about a legal process: the low-effort version explains the procedure as stated in the act, which anyone can find. The high-effort version explains where clients typically get stuck, what documentation the registrar in your city actually asks for, how long it took in the last few matters you handled, and which of the standard online explanations is out of date.

A dentist writing about a treatment: the low-effort version describes the procedure generically. The high-effort version includes your own before and after photographs, what recovery genuinely looks like for most of your patients rather than the textbook answer, and the questions people ask you in the chair that no article addresses.

An e-commerce store writing category pages: the low-effort version is a template with product names inserted. The high-effort version pulls in your own review data, your own product photography, return rate patterns you can see and competitors can’t, and honest guidance on which product suits which customer.

A local service business: instead of one page per city with the name swapped, write about jobs you actually did in those areas. Photograph them. Mention the specific conditions that make work in that locality different.

The pattern across all of these is the same. You already have the material. It’s sitting in your files, your inbox, your camera roll and your memory. The work isn’t research. It’s extraction.

Where AI Fits

This resolves more cleanly than most discussions of AI and SEO suggest.

Google isn’t penalising AI involvement. It’s assessing whether the finished page contains anything of unique value. That means the division of labour is fairly obvious.

AI is useful for structuring an argument, drafting sections you’ll rewrite, tightening prose, generating an outline, summarising your own notes, and catching gaps in coverage.

AI cannot supply the facts you learned by doing the job, the photographs of your work, your data, your judgment about which option suits which customer, or the specific thing your clients always get wrong.

Which means the risk isn’t using AI. The risk is publishing a page where the AI portion is the only portion. A page assembled entirely from what’s already on the internet is low effort regardless of what produced it, and it was low effort before AI existed too. The tools have just made it much easier to produce a lot of it quickly.

The Uncomfortable Implication

Here’s the part worth sitting with, because it cuts against how most people think about content work.

You can work extremely hard and produce low-effort content.

Spending two weeks researching a topic online, reading twenty articles and carefully synthesising them into a comprehensive guide is genuinely hard work. It’s also, by this definition, low effort, because everything on the page came from somewhere else. Nothing on it required you specifically.

Meanwhile, spending two hours writing up something you learned from six years of doing the job, with photographs from a project you completed last month, is high effort by the same standard. Less labour, more evidence.

That reframe changes what’s worth doing. The instinct when a page underperforms is usually to make it longer, cover more subtopics, add more sections. Often the better move is to remove the material anyone could have written and add the material only you could.

Shorter and more specific beats longer and more comprehensive, when the comprehensiveness is assembled from other people’s work.

One More Thing Worth Noting

Google’s Danny Sullivan has made a related point about differentiation, using the example of travel sites that all end up with the same layout, the same author sidebar and the same features. Publishers look at what’s ranking, conclude that’s what Google wants, and copy it. The result is a category of sites that are indistinguishable from each other.

His observation is that this isn’t actually what Google is asking for. Copying the winning format is a reasonable instinct and a poor strategy, because whatever made the original page good was probably specific to it.

Which lands in the same place. The point isn’t to look like a high-effort page. It’s to put something on the page that only you could put there.

Frequently Asked Questions

Does Google actually have an effort score?

The Rater Guidelines instruct human raters to assess effort, and it’s mentioned over a hundred times. Separately, Google’s leaked API documentation contains attributes named contentEffort and ugcDiscussionEffortScore. What we don’t know is whether or how those are used in live ranking. Treat the guidelines as the reliable source and the leak as supporting context.

Do longer articles rank better?

Not because of length. The guidelines treat padding as filler, meaning content that inflates a page so it appears substantial while adding nothing useful. Length that comes from genuine additional detail helps. Length added for its own sake works against you.

Will Google penalise my content for using AI?

Not for using AI. Google’s stated position is that it assesses quality and purpose, not production method. What gets treated as low effort is content with no added value over what already exists, whether a person or a tool produced it. Keep first-party material in your content and the tooling question mostly resolves itself.

How do I show evidence of first-hand experience?

Describe your methodology. Include your own photographs and screenshots. Give specific numbers from your own work. Mention edge cases and things that went wrong. Reference what clients or customers actually ask you. These are the details that can’t be researched from outside.

Do stock photos hurt my pages?

They don’t demonstrate anything. Borrowed, stock and lightly altered imagery reads as low effort in the guidelines, while original photography and video count as evidence of work. For most small businesses, ordinary photographs of real work outperform polished stock images on this measure.

What should I do with existing content that’s mostly recycled?

Audit rather than delete. For each page, ask what only you could add, then add it. Often that means cutting the generic opening section and replacing it with something specific. Improving a handful of pages properly tends to beat publishing more of the same.

Does this apply to service pages and product pages, or only blog posts?

All of them. Product pages benefit from your own images and real review data. Service pages benefit from specifics about how you actually work. The contentEffort attribute is described as relating to article pages, but the guidelines discuss effort across page types.

Is discussion or opinion always needed?

No. Some pages are purely factual and adding opinion would be padding. But for anything where the reader is deciding between options, a judgment based on evidence you’ve considered is a large part of what makes the page worth reading.

If You’re Working Out Where to Start

The framework is straightforward. Applying it across an existing site is less so.

Working out which pages are worth improving rather than replacing, what first-party material you already have but haven’t published, how to extract it from people who are busy doing the actual job, and how to build a process that keeps producing it rather than reverting to generic content after two months. That’s the difficult part, and it’s mostly organisational rather than technical.

It’s the core of the work I do with clients on SEO and content: finding what a business already knows that nobody else can claim, and getting it onto pages in a form that both people and search engines can recognise as worth something.

If you’ve been publishing consistently without much to show for it, this is usually where the problem sits, and it’s worth a proper look rather than another round of keyword research.

Click here and tell me what you’re working on

Happy to look at a few of your pages and tell you honestly what’s missing.

This article is for informational purposes only.

Deepak Vankar is a digital marketing consultant working with businesses and professionals on SEO, content, performance marketing and brand strategy.