YT YouTube2Text

YouTube2Text → Transcripts → Google Just Defined "Low-Effort" Content (Big Deal for SEO)

Google Just Defined "Low-Effort" Content (Big Deal for SEO) — Transcript

by Matt Kenyon · 3,890 words · 570 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00So, Google just updated their helpful
0:02content guide and added some very
0:04interesting language that's leading a
0:06lot of people to think that something
0:08pretty big is on the horizon with
0:10Google's algorithm, specifically
0:12targeting AI content. So, if you use AI
0:16in any form of your content production
0:18pipeline, this video is for you. See,
0:20most people think AI content is safe as
0:22long as a human reviews it before it
0:25goes live. But in her recent article,
0:27Lily Ray looked at sites Google hit this
0:30year, and several of them said they had
0:32human editors checking every draft. Yet,
0:35Google demoted them anyway. As I said,
0:37on October 1st, Google changed its
0:39helpful content guide, and it laid out
0:42exactly what it's looking for, and we
0:43will go over that criteria in a second.
0:46But also, the fact that they updated it
0:48at all is a sign that something big may
0:50be coming, and I'll talk about that as
0:52well. Google actually updated two pages
0:54that day. The first is its guide to
0:56creating helpful content. It added a new
0:59category that a lot of people have been
1:01talking about online called main
1:03content. And main content is basically
1:05anything on the page that helps it do
1:07its job, like the text, images, videos,
1:10tools, reviews, and even the headings.
1:12Then it also lists the four things that
1:15its quality raers judge that content on,
1:18which is effort, originality, talent or
1:21skill, and accuracy. And if you don't
1:23know what quality raiders are, I'll
1:25explain more why that's so important. It
1:27also says that making up authors with AI
1:29head shot, fake names, or fake
1:32credentials is officially a form of
1:34deception. And yes, fake authors have
1:37actually been something very popular
1:39SEOs have recommended in the past,
1:41especially on affiliate sites, which
1:43feels absolutely insane in hindsight.
1:46So, I'm glad Google has unambiguously
1:48condemned that so that we can finally
1:50just put that behind us. The second is
1:53Google's guidance on AI content. It now
1:56says it's quote critical to factcheck
1:59and review all AI generated content
2:02before you publish it. And that includes
2:05your titles, metadescriptions, and image
2:07alt text. Now, this is probably pretty
2:09common sense, like of course you want to
2:10review content, but the fact that Google
2:12mentioned this specifically points to
2:15something much bigger, which I'll talk
2:16about in a second. And we will get into
2:18more of what that fact check and review
2:20process should look like to protect your
2:22site against getting penalized. But
2:24first, I want to talk about who the
2:25quality raiders are that I just
2:28mentioned. That may be a new concept for
2:29you. See, Google pays thousands of
2:32people, yes, real humans, not algorithms
2:35or AI models, if you can believe it, to
2:38look at search results and score how
2:40good they are using a long set of
2:43guidelines. And Google makes no secrets
2:44about this. They even publish the
2:46guidelines themselves. They also grade
2:48the ranking changes that Google tests
2:51before launching them. So, Google says
2:53raiders have no control over how pages
2:56rank, and that is true. It's not like
2:58raiders are, you know, looking at your
3:00page and I'm going to move this one down
3:01and move this one up manually. That's
3:03not how it works at all. Raiders don't
3:05actually rank your page. They train the
3:08systems that do, if that makes sense.
3:11So, when Google tells raiders to judge
3:13effort, originality, skill, and
3:15accuracy, it's telling you what its
3:17ranking systems learn to reward. So, why
3:20does this matter right now? Well, before
3:23I tell you, if you're interested in
3:24content like this, where I help you win
3:26in search, build with AI, and grow with
3:28video, please subscribe to the channel.
3:29And while you're down there, give this
3:31video a like. It means a lot, and I
3:32really appreciate it. Anyway, this is
3:34why it matters because Google has a
3:37habit of telling us what's coming before
3:40it hits. Lily Ray laid out the pattern
3:42in her article, which I highly recommend
3:45checking out. The link will be in the
3:46description. In January 2011, Google's
3:49Matt Cuts wrote that people were asking
3:52for stronger action on content farms.
3:55The next month, Google released Panda.
3:58In March 2012, he said Google was
4:00working on a penalty for overoptimized
4:02sites. The next month, they released
4:04Penguin. Now, these are the names of the
4:06Google updates, very big Google updates.
4:08Now, in July 2018, Google updated its
4:11raider guidelines to focus on
4:13reputation. And what happened 12 days
4:15later? The medic update hit and this
4:18year on May 15th, Google clarified that
4:20its spam policies apply to AI overviews
4:23and six days later, you guessed it, a
4:26core update rolled out. Need I continue?
4:29The other signs are stacking up, too.
4:31Google has run four spam updates this
4:33year compared to one in all of 2025, and
4:36the gaps between them keep getting
4:38shorter and shorter. And at a Google
4:41event in Barcelona, Google's Gary Ilash
4:44said, "Scaled content is becoming a
4:46bigger problem than link spam." Whoa.
4:49Never thought I'd see the day where
4:50something would become even spamier than
4:53spammy backlinks. Now, nobody outside
4:55Google knows when the next big update
4:58lands or whether it'll be its own named
5:00update or part of a core update. But, I
5:02mean, look at the pattern. If history
5:04repeats itself, the October 1st rewrite
5:07is a pretty significant warning that
5:09something big is coming. Specifically,
5:12something targeting AI content. Okay, so
5:15let's go through the four tests one at a
5:17time. These are tests straight from
5:19Google that it's saying it applies to
5:21your content to know whether or not to
5:24rank it highly. The first is effort.
5:27Google defines this as how much human
5:29work went into creating the content or
5:32the systems powering it. Its examples of
5:34high effort are writing original
5:37analysis, translating a poem by hand,
5:40and building a custom interactive map
5:42for a game. Very niche examples there.
5:45Its examples of little to no effort are
5:48pages generated automatically from feeds
5:50and using AI to produce large amounts of
5:53text without manual oversight or
5:56curation. Google also says that citing
5:58your sources does not replace the need
6:01for original effort. Now obviously
6:04nobody can see how much effort you put
6:06into a piece except you, not even the
6:09quality raers. What they can do though
6:11is look for signs of low effort. So to
6:14me that means general knowledge that any
6:17LLM could spit out with zero original
6:20thought and zero firsthand experience. A
6:23term that's become very popular online
6:25is a meat proxy. Essentially you're just
6:28giving the query to the large language
6:31model saying write a blog post on this
6:33and then passing it back to the user. So
6:35you're not adding anything. You are a
6:37meat vessel for the large language model
6:40to just give information a wall of text
6:42to the user. And it's like if they
6:43wanted to consult a large language
6:45model, they'd consult a large language
6:47model. They're going to your website
6:48because they want to hear your expertise
6:50and your opinion. And if you spend a lot
6:51of time online, you can usually tell
6:54what this loweffort AI generated content
6:56looks like. I mean, there are the
6:58obvious tells like, you know, insert
7:00joke about M dashes here. And of course,
7:02the it's not X, it's Y. But there are
7:04deeper ones, too. Things that go beyond
7:06just like surface level semantics. Like
7:09large language models tend to really
7:11hedge. They don't often have strong
7:13opinions. Their writing is almost too
7:16structured, like too perfect and built
7:18in a way that no human actually talks to
7:21another human. It's hard to describe,
7:22but you know it when you see it. They
7:24also tend to circle an idea a lot
7:26instead of just saying it clearly. So,
7:29we all kind of know these patterns
7:31intuitively, and they're exactly what I
7:33look for and sort of what my brain has
7:36been trained for whenever I'm reading
7:38content online. And that leads to the
7:40second test, which is originality. So,
7:43does your content offer information or
7:45perspectives that aren't already on
7:47other [snorts] websites? So, if your
7:49article is literally just a remix of the
7:51top 10 results, it's going to fail. I
7:53mean, your own data, your own tests,
7:56your own opinion, your own experience,
7:59that's what's going to pass the test
8:00here. A great example of this is a site
8:02called Crutchfield. So, Crutchfield is
8:05like an AV distributor and they sell a
8:08lot of car audio supplies. They sell a
8:10lot of home audio supplies and they do a
8:13phenomenal job of just putting so much
8:16effort and originality and care into all
8:19their content that I think they're a
8:20great example. So, you have this article
8:22here, which is honestly one of their top
8:25ranking articles, how to wall mount your
8:28TV, a step-by-step guide, and you
8:30literally have a picture of the guy
8:31writing the article here. And then you
8:33have a video of him literally in a
8:35warehouse showing you how to wall mount
8:37a TV with this like, you know, setup
8:40drywall here. Like you can't get more
8:42detailed than this. He's giving you the
8:44preparations. It's actual screenshots of
8:47equipment that you need that they're
8:48taking. They got custom graphics here
8:51telling you like how high you want to
8:53have your TV, how to center it at your
8:54eye level. And this is all of their
8:57content. So all of their content is
8:59super original, super high effort.
9:02There's original photos that honestly
9:04don't even need to be that fancy. Like,
9:05you could take these photos with your
9:07iPhone. You got a step-by-step process
9:09here showing you everything. And like I
9:11said, there's, of course, the video that
9:13shows it in action. Here's another
9:15article about how to wire in your
9:17subwoofers. And again, as you would
9:19imagine, they have like all these
9:21superdetailed diagrams. They have images
9:24of their staff actually wiring up a real
9:26subwoofer. Here you have these drop-
9:28down things where you can basically have
9:30all the wiring diagrams that you need
9:33based on the amplifier that you have. So
9:35they really have thought of everything.
9:36If I have a mono one channel amplifier
9:38and three subwoofers, I can click this
9:39and get the accurate diagram. So it's
9:42just you could tell that this is
9:44whatever the opposite of AI slop is.
9:46This is what Crutchfield is doing.
9:47They're just pouring so much love and
9:49care and effort into this content. And I
9:51think that's going to be a huge
9:53differentiating factor and I think
9:54honestly a big reason why they win the
9:57SERs for their categories is because of
9:59content like this. Another example in
10:01the SAS world is ReLA. So Reilla makes
10:04software that records in-person sales
10:05appointments like a home improvement
10:07salesperson sitting in someone's living
10:09room and it uses AI to coach the reps.
10:12Its research team Labs publishes studies
10:16from that data. In one, they looked at
10:1910,000 recorded appointments in
10:21two-party consent states where everyone
10:24has to agree to be recorded. And they
10:26extrapolated some really cool firstparty
10:29insights from that data. That's the kind
10:31of thing that a large language model
10:33simply can't replicate. So, the next
10:34criteria is an interesting one, and it's
10:37talent or skill. So, what in the world
10:39does that mean? Well, Google's examples
10:41are clear writing, a well-produced video
10:44and page tools that actually work. So,
10:47by the way, here we have Google
10:49acknowledging the power of embedded
10:50tools. I made a video on that like a
10:52year, a year and a half ago, how you can
10:54make your own tools with cursor, even if
10:56you're not technical. And a large
10:58language models have of course come so
11:00far since then. So, you can check that
11:02out if you'd like. Needless to say, I'm
11:03a huge fan of embedded tools in content,
11:07especially if they help the user solve
11:09their problem. And interestingly, Google
11:11says not everything needs an expert. So
11:14its example is someone sharing a new way
11:17to clear snow off their driveway, which
11:19is another very specific and very niche
11:22example. That's something that you don't
11:24need credentials for. You just need real
11:26experience and the ability to replicate
11:28that experience. And the fourth
11:30criteria, the fourth test is accuracy.
11:32So, pretty common sense here, but with
11:34the proliferation of scaled AI content,
11:37it bears repeating. For informational
11:39pages, the facts have to be right. For
11:42topics that can affect someone's health,
11:44money, or safety, Google calls these
11:46your money, or your life topics, the bar
11:48is especially high. The content must be
11:50highly accurate and match expert
11:53consensus. So, how do these four
11:55criteria affect the AI scaled content
11:58that Google seems to be targeting here?
12:00Well, in my opinion, effort by Google's
12:03own definition because scaling AI
12:06content without any human oversight
12:08takes very little effort. And I think
12:11originality, too, because a model
12:13trained on the web tends to repeat
12:15what's already on the web. See, if I
12:18were to read between the lines here,
12:19which a lot of people have already done,
12:21I think what Google's saying is, guys,
12:24you got to stop just like taking a user
12:26query and throwing it into chat GPT and
12:29saying, "Write me a blog post on this
12:31query and then copying that text and
12:34pasting it into your website without
12:35changing anything." Like, guys, can we
12:37please not? Like, I think that's kind of
12:40what's what's happening uh under the
12:42surface there. So looking at sites that
12:45got hit this year, Lily Ray found a few
12:47clear patterns for sites that got
12:49slammed by algorithm updates. The first
12:52pattern is scaled pages. That's hundreds
12:55of thousands of pages published in a
12:57short window, often on topics the site
13:00has no real experience in. That includes
13:03templated pages, so like one page per
13:05city or per product comparison with only
13:08a few words changed between them. Sites
13:11doing this have been getting hit for at
13:12least 15 years. AI just sort of made it
13:15easy to speedrun their demise. The
13:17second is fake authors. So, I mentioned
13:19this one already. If the author doesn't
13:21exist, Google now calls it deception in
13:23writing. It's a dumb idea anyway. Don't
13:25even entertain it. Next,
13:27self-promotional listicles. Oo. Okay.
13:29So, that's a best X for Y type list that
13:33a company would publish on their own
13:35site and they would put their product at
13:37the top. I've thought a lot about this.
13:38My last video was all about this. Now,
13:40on this one, I'm not exactly sold that
13:45listicles or self-promotional listicles
13:47in general should be completely thrown
13:49out. I'm kind of waiting to see what
13:52happens here. The data is a little
13:54complicated. So, I ran my own study on
13:56this in March. I tested 116 best CRM
14:00listicles written by CRM brands in Chat
14:03GBT. And only about a quarter of the
14:05time did the brand that wrote the
14:07listicle come out as the top pick. And
14:09over a third of the time, Chad GBT
14:11actually cited the brand's listicle and
14:13recommended only its competitors. In
14:15other words, a lot of times these
14:17listicles completely backfired and
14:19actually made things worse for the
14:21brand's AI search profile. That said, a
14:24lot of smart, wellrespected SEO
14:26researchers say that these still work.
14:29So, I'm not calling them bunk
14:32completely, but at this point, I'm not
14:34recommending them to clients right now
14:36either. I'm I'm basically like Kurt
14:38Russell at the end of the thing, just
14:39like sitting there in the snow like,
14:41let's just wait here a while, see what
14:43happens. The reason I struggle with this
14:45is because I think it was Rand Fishkin
14:47or someone who said like it always stuck
14:49with me, you should produce content like
14:51search engines don't exist, right? And
14:53and just to be as helpful as humanly
14:55possible. And honestly, if I'm thinking
14:58from that mindset of just like I'm going
14:59to be a consultative salesperson to my
15:02ideal buyer, yeah, I would maybe write a
15:04listical and I would mention
15:06competitors. And I think this listical
15:07thing can only work if it's not
15:10blatantly self-promotional. So, for
15:13example, let's say I'm writing a
15:15listicle on the best digital pianos for
15:19beginners. Like, I'm thinking of that
15:20cuz I have the Yamaha P45 right here.
15:23And in that listicle, I'm saying all the
15:26things that all these digital pianos are
15:27really good at for beginners. But then
15:29maybe I'll say something like, "Hey, if
15:30you want a full-size piano, then the
15:32Yamaha P45 is great because it's got all
15:3488 keys. However, if you want something
15:36portable, it's not a good idea because
15:39this thing is like a bear to try to move
15:41around. So, in that way, I am mentioning
15:44competitors and I'm being as helpful as
15:46possible, but I'm doing so factually and
15:48with integrity. I think the
15:50self-promotionalistic that Lily's
15:51talking about and maybe I don't know I
15:53haven't asked her about this are things
15:54that are just blatantly and sort of
15:56obnoxiously self-promotional where
15:58basically these people are just trying
16:00to game the system by putting their
16:02solution first and then listing out a
16:04bunch of competitors and just saying
16:06like a bunch of either untrue or not
16:08nice or generic or you know whatever
16:11shady things about their competitor. I
16:13think that obviously is a very sketchy
16:15tactic and I think that deserves to die.
16:18But I'm not 100% sold that listicles in
16:21general need to go away forever because
16:24if I'm thinking from a prospecting and a
16:26qualification standpoint and I'm selling
16:28like these keyboards, I don't want to
16:30sell it to someone who wants something
16:32portable. I want to make sure that
16:33there's good like product market fit and
16:36good buyer product fit, those sorts of
16:38things. So, I kind of want to show them
16:40some alternative routes and hopefully
16:42build their trust in that way. Anyway,
16:44let me know what you think about
16:45self-promotional listicles. Drop a
16:46comment. I'd really be interested in
16:47hearing that. So, something that's
16:49actually kind of crazy and pretty funny
16:51is that Chad GBT is actually starting to
16:53call brands out for their
16:55self-promotional listicles. Uh Lily
16:57posted this just two days ago. She said
16:59another Chat GBT disclaimer about
17:01self-serving rankings in B2B. And it
17:03says someone searched basically for a
17:05for a a particular product. And it said
17:08one important caveat. A lot of best
17:09membership software rankings are
17:11extremely selfserving. Vendors
17:14frequently publish comparison pages that
17:15conveniently rank themselves highly. I
17:18weighted recent third party capter data
17:20more heavily than vendor listicles for
17:22that reason. So clearly Chad GPT is like
17:25we're not having none of this. This was
17:26a fun exploit while it lasted. But I do
17:29think that they're going to be cracking
17:30down on this more and more. According to
17:32Lily's research updated in September,
17:34Google's AI overviews now site 38% fewer
17:37of these listicles, and they leave the
17:39publishing brand out of the
17:40recommendation 83% of the time. Another
17:44area that I think is going to get
17:45absolutely slammed is paid mentions. So,
17:48there's a growing market selling spots
17:50in other sites best of lists, often like
17:53$135 [snorts]
17:54to $250 each, just to try to get brands
17:57into AI answers. If these spots come
18:00with links that pass ranking credit,
18:02that is technically buying links, which
18:03Google has penalized for years, I think
18:06they're going to pay really close
18:07attention to these sites now as they try
18:08to clamp down on this. Again, it's going
18:10to be really hard for Google to prove
18:12that you bought those mentions or links.
18:14So, my guess is that Google will start
18:16to weed out these mention reselling
18:18domains and basically just destroy any
18:20authority associated with them.
18:22Basically, if something looks sketchy
18:23and they're asking you for a paid
18:24placement, probably best to stay away. I
18:26mean, so listen, I love AI. I genuinely
18:28could not run my business without it.
18:30And so, we have to talk through how to
18:32use AI responsibly without ending up on
18:35Google's hit list. Here's what I would
18:37do. First, and probably most
18:39importantly, I would start with
18:41something that AI does not have. So,
18:43something that's just not in the
18:45training data. So, your own data, a
18:47customer call, a test that you ran,
18:50maybe even like a strong opinion that
18:52goes against the industry consensus.
18:54Give that to the AI and let it help you
18:56shape it. Let it help you steal man the
18:59other side and poke holes in your
19:00argument and make it even stronger.
19:02Next, factch checkck everything. Check
19:05every number. Quote, and claim against
19:08the original source and open every link
19:10by hand. Do not delegate all of this to
19:13AI. Highly recommend you go in and
19:16review all claims, all arguments, all
19:18first party sources. Google called that
19:21review quote unquote critical, so treat
19:23it that way. And I can't believe I need
19:24to say this, but you or a trusted editor
19:27must read every word of your AI output.
19:30Look for the tells I mentioned earlier,
19:32like hedging, no real opinion, and
19:34circling the point. If it sounds like an
19:37LLM, rewrite it until it sounds like
19:39you. Next, put a real name on it. So, a
19:43real author with a real bio. So, for me,
19:45it wouldn't be written by the Kenyon
19:47digital team, but written by Matt Kenyon
19:50with a clickable author bio. And if AI
19:52did a big part of the work, say how you
19:54used it. Honestly, yes. Actually say, "I
19:58used AI to help me with this, and here's
20:00how." I for me, I feel like you'd earn
20:03my respect more, and I think you'd earn
20:05your audience's respect more if you just
20:07admitted it because I mean, AI itself
20:09isn't penalized in Google's eyes. It's
20:11more so the loweffort form of AI content
20:13that we've been talking about. And
20:15lastly, don't scale past your expertise.
20:18So I see this phenomenon with
20:20programmers, with knowledge workers,
20:21with marketers, everyone who uses AI. It
20:24is so tempting to just get lazy and let
20:27the AI run ahead of you and just hit yes
20:30to all, accept all changes, not even
20:33read anything. And honestly, for the AI
20:35to sort of go beyond your own
20:37understanding of a topic and I think
20:39that's where you can get in big trouble.
20:40So that's the list. Effort, originality,
20:43skill, and accuracy. Google hasn't said
20:45when the next update lands, but if
20:46history repeats itself, I think it's
20:48coming soon. I think it's coming soon.
20:50Let me know if you're watching this in
20:52the future. Comment below or maybe I
20:53will if the update comes. You can check
20:55out all the articles, exposts,
20:58everything below on all the really smart
21:00people that have been talking about this
21:01stuff. And give me a follow on X, guys.
21:03I mean, I'm struggling over there. Your
21:05boy's struggling. Feels like I'm just
21:06tweeting into the void. Anyway, thanks
21:09for watching and happy ranking.

About this transcript

This page contains the full transcript of Google Just Defined "Low-Effort" Content (Big Deal for SEO) by Matt Kenyon, generated from the public captions YouTube serves with the video. The transcript has 3,890 words across 570 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.