Google Is Killing Pages (Here's How to Fix It)

0:00Google reached into the index, found

0:01pages that had been sitting there

0:02ranking for months, and pulled them back

0:04out again, flipped them to crawled,

0:07currently not indexed. About a third of

0:09the location pages on one client site

0:11gone in a couple of weeks, and nothing

0:13on those pages had changes. So, when I

0:15went looking at what sites this was

0:17hitting, there was a clear pattern to

0:19this. A site with two or three of these

0:21pages was fine. Nothing happened. A site

0:23with 30 or more, they were losing most

0:26of them. Now, one system generated all

0:29of them to the same spec at the same

0:30quality. The only variable was how many

0:33each site had, which means Google is

0:36grading these pages against each other

0:38on the same domain. And somewhere in

0:40there is a point where the next page you

0:42publish stops helping you and starts

0:45costing you. Now, I found where that

0:46point is and what it takes to get back

0:49across it. It took a human copywriter,

0:52yes they do still exist, three pages and

0:54about 11 hours of research. Now, first

0:57the timing because this got reported

0:59badly everywhere. There was no April

1:01core update. Google confirmed to this

1:03year, March 27th through April 8th and

1:06May 21st through June 2nd. What hit us

1:08landed in the gap between in early May.

1:11And it wasn't only my client's sites.

1:14Starting in late April, site owners

1:16everywhere were reporting the same

1:18thing. pages [snorts] that have been

1:19indexed for years. No manual action, no

1:22crawl error, nothing broken, suddenly

1:24flipping into crawled, currently not

1:26indexed, and staying there. John Miller

1:29got asked about this directly on April

1:3130th. His answer was that super helpful.

1:35Some sites go up, some sites go down,

1:37and he didn't see anything exceptional.

1:39Now, if you were watching your pages

1:41disappear that week, that landed about

1:44as well as you'd expect it to. Now,

1:46here's the part that not a lot of people

1:48understand, but it changes how you read

1:51your search console. Crawled, currently

1:53not indexed, has nothing to do with

1:55ranking. It happens in a different part

1:58of Google's system before ranking is

2:01really in the picture at all. A core

2:03update moves around pages that are

2:05already in the index. This is Google

2:07looking at a page and deciding it isn't

2:11worth the storage. Storage costs money.

2:13Google crawls far more of the internet

2:15than it ever intends to keep. And every

2:18page it keeps is a page it pays to hold.

2:21So a page in that bucket got read and

2:25turned down later. This is a harsher

2:28judgment than a ranking drop. A ranking

2:30drop says you're not the best answer

2:32today. This it says the page doesn't

2:35need to exist. No user of Google

2:39anywhere on the planet ever is going to

2:41find value in the content on that page.

2:43Therefore, Google won't index it.

2:45Miller's been clear about what drives

2:47it. He said if their systems are

2:49seriously worried about the quality of

2:50the website, they crawl it less and

2:52index less. And this status is what that

2:55looks like from your side. Gary Illis

2:57said the same thing from the other

2:59direction. That general site quality

3:01matters a lot for how many of these you

3:03see. So the status is a readout of

3:05something Google already decided. And

3:08this left us with location pages Google

3:10had accepted years ago and then rejected

3:13again. Nothing had changed on our end.

3:15So something changed in what Google was

3:17willing to hold on to. And the pages

3:19closest to the line were the ones that

3:22fell off it. The only way to find out

3:23which line was to take a few of them and

3:26change them. So we pulled three of the

3:28drop pages and I handed them to my human

3:30copywriter. We didn't touch anything

3:32else. We didn't source links. We didn't

3:34do schema. We didn't do anything

3:35technical. Same URL, same template.

3:36Everything was kept the same. The only

3:38variable was the content, the words on

3:41the page. He spent about three and a

3:43half hours on each one of them. And what

3:46he did with that time is the whole

3:48answer of what Google is looking for.

3:50So, let me walk through. Okay. He

3:52started in the search console and in the

3:54Google business profile performance

3:55insights looking for what Google already

3:58associated that page with. looking for

4:00what search queries that page was

4:02already indexed. He wasn't after the

4:05keywords that he liked or that he

4:06thought it should be indexed for. He

4:08wanted the queries the page was already

4:10surfacing for, even weekly, even ranked

4:13in position 95, because that tells you

4:15what Google is already thinking the page

4:18is about. You're adding onto something

4:20instead of arguing with Google's

4:21algorithm. Don't argue with machines. It

4:24rarely works out. Then he went through

4:26the client's call transcripts, actual

4:28recorded calls from actual customers,

4:30looking for what people say out loud

4:32when they call the business owner. This

4:34is different from how they type. And

4:36this exists nowhere else on the

4:38internet. Then he went to Reddit, not

4:40the topic in general, but that specific

4:43neighborhood, what people who live there

4:45complain about and warn each other

4:46about. This client is in Chicago, so

4:48there's a lot to work with. And then the

4:51piece that mattered the most. He took

4:53these pages and turned them into a

4:55story. An actual throughine connected to

4:57the business, its history, and why

4:59someone in that part of Chicago ends up

5:02calling their office. Most location

5:04pages are a pile of generic neighborhood

5:07facts with the service bolted on to the

5:08end. This was not what he wrote. Three

5:11pages, 11 hours of work. All three got

5:15indexed within a week. Now, three pages

5:17is an anecdote. And yes, I'll say that

5:19plainly. We try to use real data in

5:22statistically significant samples before

5:24we make any conclusions. But it told us

5:26the failure was likely in the words and

5:28we finally had a description of what had

5:31been missing. So we took what the human

5:33did and rebuilt our content generation

5:36process around it. The research order,

5:38where the data comes from, how the

5:39outline gets built before anything gets

5:41ever written. Across every site we run

5:43it on, we now see location pages

5:45indexing at about 97%.

5:48They were somewhere between 60 and 70%

5:51before this change with Google's update.

5:53And that range is wide for a reason,

5:56which brings me back to what I said at

5:57the start of this about volume. The 60

6:00to 70 was never a flat rate. It depends

6:02almost entirely on how many of these

6:04pages a site already has. A site with

6:08two or three location pages, yeah, they

6:10indexed almost every time. A site with

6:1230, most of them get rejected because

6:15they're built the same way to the same

6:17spec. the count was the only difference.

6:19And once you understand what Google is

6:22actually deciding, it starts to make

6:23sense. Miller framed this in a way that

6:26I haven't been able to stop thinking

6:27about. He said, "The question isn't

6:29whether the page is good. Google's

6:31algorithm is very bad at deciding

6:34whether content is quality or not. The

6:37question is what is missing from the

6:39index if this page isn't in it." a

6:42solidly written article that already

6:44exists in a similar form 30 times over

6:47doesn't say anything new. And he took it

6:49further and said about AI content

6:51directly that you read some of these

6:53pages and you think, boy, anyone could

6:55have written that. That's how AI writers

6:57typically work. They produce the average

6:59of what already exists. So page one

7:02clears easily. Nothing else in the index

7:04covers that service in that neighborhood

7:06from that business. Something is missing

7:08without it. But by the time you get to

7:09page 30, you've answered that question

7:1129 other times, each page is competing

7:14against your own previous 29. And that

7:17they're all built the same way with the

7:19same depth. The honest answer to what's

7:20missing without this 30th article is not

7:24very much. And that's the threshold. It

7:26isn't a number Google publishes, and

7:28it's likely different for every site,

7:30for different niches and different

7:31areas. But the mechanism is the same

7:33everywhere. Every page you publish

7:35raises the bar for the next one because

7:37the next one has to be worth keeping in

7:39a world where your previous pages

7:41already exist. So volume on its own is

7:43fine. It turns on you when the pages

7:45stop getting more specific as you add

7:48them. All right, I want to show you what

7:49this looks like in the search console

7:51because I want to be careful about the

7:52difference between the number moving and

7:54the number meaning something. So this is

7:56the coverage report on the client site

7:58for crawled currently not indexed. This

8:00is the category where Google looked at

8:02the content on the page and decided it

8:04never needed to show it to anyone ever

8:06in the world. Okay. Now, this is a real

8:10business. So, 277 pages sitting in

8:13crawled not indexed. But we can see this

8:15peaked at 363 at the end of June. Now,

8:20watch how it moves through steps rather

8:21than slopes. One step down here,

8:25one step down here. This is Google

8:27reprocessing, rec crawling this website

8:29in batches, which is what you'd expect

8:31when a group of pages gets rewritten and

8:33recrolled all at once. Okay, now I want

8:35to be honest about these 277 pages. Most

8:38of them are not location pages. Most of

8:40this 277, something like 87% of them are

8:45machine translated Spanish pages, plus a

8:47bunch of duplicate junk, tracking

8:49parameters, HTTP versions of HTTPS,

8:51trailing/variants, stuff like this. None

8:54of this was ever really meant to be

8:55indexed. So, I don't really count it.

8:56The number that I actually watch is

8:58nine. Right now, there are nine location

9:01pages de-indexed. That's compared to 89

9:04that were de-indexed at the end of June.

9:06And two of those nine are duplicates,

9:08which we should fix rather than real

9:10failures. So, this is seven out of 89

9:13that are not indexed. That's 90%. This

9:16is below the 97% average, but I'm

9:18showing you this below average one on

9:20purpose because honestly, anyone can

9:21screenshot their best client. Now, those

9:23seven were last crawled between July 2nd

9:26and July 31st, and the rewrite that we

9:28published went live after that. This

9:30means Google hasn't seen the new version

9:32yet, and I expect them to be indexed

9:35when Google does. If they don't, I'll

9:37record another video and let you know

9:38why. But there's a second number on this

9:40report that almost matters more, and

9:43very few people check it. Let me show

9:45you. This one is discovered, currently

9:48not indexed. Okay, five pages on a site

9:51with hundreds of URLs, hundreds of

9:53pages. This is what tells me that Google

9:55sees this as a healthy domain because

9:58those two statuses answer different

9:59questions. Crawl, not indexed, means

10:01Google came, it read, and passed. A

10:05judgment about the page. A judgment

10:06about the quality. Discovered, not

10:08indexed, means Google knows the URL

10:10exists, but hasn't even bothered to go

10:13look at it. Okay, that one's about you.

10:16Google's crawling documentation explains

10:18why. They call it crawl demand, and it

10:21reflects how interested Google is in

10:22your pages, including how good it thinks

10:26they're going to be. Low demand means

10:28less crawling, no matter how fast your

10:30server is, and their own guidance.

10:32There's a lot of low-v value URLs will

10:35hurt a site's crawling and indexing. So,

10:38crawling first in that sentence, right?

10:40So, there's a progression available

10:42here. A template keeps producing pages

10:44that Google won't keep. that results in

10:46Google's interest in that pattern

10:48dropping and eventually the new one stop

10:50getting crawled at all. Now, Google

10:53hasn't laid this out as a formal

10:55sequence, so call it my read, my

10:56educated guess, but this lines up with

10:58everything they published about crawl

11:00demand. So, on this site, the number is

11:02five. Google is still interested.

11:03They're crawling almost every new page

11:05we publish. It's just turning down these

11:07specific five pages. Okay, these pages

11:10have a problem. The site doesn't. So,

11:12what do you actually do with a page

11:14that's in this bucket? Because there's a

11:16way to drive this number to zero that

11:18really fixes nothing at all. And it's

11:20the first move a lot of people reach for

11:22the magical no index tag. Page shows up,

11:26it's on one of these reports, you don't

11:27want it, throw a no index tag on it, and

11:29it drops off the report. Number goes

11:31down, dashboard looks clean, and you

11:33have changed absolutely nothing. Those

11:35pages are already not indexed. That's

11:38what the status means. The tag moves

11:40them from one bucket in your report to a

11:42different bucket in your report while

11:44Google's opinion of your site sits

11:45exactly where it was. You've

11:47disconnected the gauge from the engine

11:49and then congratulated yourself on the

11:52reading. Okay. Now, forcing this also

11:54doesn't work. Resubmitting, pointing

11:56internal links at it, pinging the site

11:58map. Miller's been direct that this

12:00isn't a technical problem to be fixed.

12:03The question Google asked was, "What's

12:06missing from the index without this

12:08page?" and submitting it again doesn't

12:11change the answer. You have two real

12:13options here. Number one, improve it so

12:16it carries something that could only

12:18have come from this business that adds

12:21meaningful information to the search

12:23results or delete the page. Deleting

12:26this took me a little while to accept.

12:28Google's own guidance on unhelpful

12:30content says removing it can help the

12:31rankings of your other content because

12:33this signal is sitewide. their words,

12:36content on a site with a lot of

12:38unhelpful content is less likely to do

12:40well, even the good content. So, a page

12:43that was never going to earn its place

12:44is costing you by existing, and taking

12:47it down is a repair of your site. Lily

12:50Ray did a study of 220 sites that scaled

12:53AI content, and it shows this from the

12:54other end. The brands that cut their

12:56content footprints are the ones where

12:58the traffic started recovering. And if a

13:00page still won't index after you've done

13:02the work, if after you've done the

13:03improvement, then the problem moved.

13:05It's sitting in whatever produced it.

13:07One page failing is a page problem. Six

13:10pages failing from the same generator is

13:13the generator's problem. So you should

13:15count rate per template instead of pages

13:17per site. Every page came out of some

13:20process, a prompt, a build sheet, a

13:22writer following a brief. Maybe maybe

13:24you still use human writers like I do.

13:27Group your unindexed pages by what made

13:29them, how they were created, and watch

13:31that rate. The level barely matters. The

13:33slope does. A template that's steady at

13:368% is noise. But if that template jumps

13:38to 40% after you change something,

13:40that's Google telling you that change

13:42made it worse. And it's telling you

13:44before your rankings move. This is the

13:47whole value. It's early. Rankings lag.

13:49And by the time a position drops, you've

13:52already published another 100 pages off

13:54of a broken template. Let me close the

13:55loop on something because I called this

13:57a leading indicator and I want to tell

14:00you about that. Every site in Ray's

14:02study, as far as we know, they indexed

14:04fine. All 220 indexed rank, pulled

14:06traffic for 6 to 12 months, then got

14:08demoted. Indexation never caught it. So,

14:10if you walk out of here thinking that

14:12your coverage report is a safety net,

14:14you're you've got it wrong. And I'd

14:16rather say that than let you find out

14:18the hard way. What this catches is thin

14:21and redundant at the door before you've

14:23built anything on top of it. Cheap,

14:26real, more than most people are

14:28checking. It's a floor. What actually

14:30kills site is slower and it doesn't

14:33always show up in search console until

14:35the traffic's already gone. So, this is

14:37why the standard I hold my own clients

14:40to lives outside the report entirely.

14:42Could a competitor publish a near

14:44identical version of this page tomorrow

14:46using the same methodology? If they

14:49could, the page doesn't need to exist.

14:51And Google will work it out eventually

14:53whether or not it indexed today. Now,

14:55everything I've covered here happens

14:57after Google reads your page. There's a

15:00whole set of problems that can happen

15:02before that where Google can't read the

15:04site properly in the first place. And

15:07those are the ones that make an AI built

15:09website invisible from day one. That's

15:12this video on the screen. Go check that

15:14one out