Now…

I previously wrote about how I built two identical websites for a real business, blocked one site from Google and one from LLMs, and how quickly I was able to get LLMs to surface and recommend the non-Google-indexed site by building a handful of links and local citations to it.

Another test I wanted to run with one of these sites was to see how quickly I could get the Google-indexed version of the site to rank for a keyword that isn’t mentioned on the website anywhere exclusively using backlinks and exact match anchor text.

To give a quick recap of the site I’m working with: I built an extremely barebones website for an established business which has been operating for over 15 years but has never had a website. This is a home and autoglass service business in a small town of about 30,000 people whose primary web presence is a Facebook page, Google Business profile as well as a lesser-used Instagram profile.

There were some parts of this that were a little tricky because although they had never had a website, they do have a significant local presence as well as the most reviews of any similar business in town so I needed to be careful to pick a keyword that was simultaneously relevant to their business but not something that gets mentioned alongside their brand frequently. I wanted to be sure that the links, specifically the anchor texts, I’m building were moving the needle and not Google simply entity matching with relevant terms mentioned elsewhere around the web. In other words, how could I be sure the anchor text was driving anything and not the domain simply gaining more authority and in turn ranking for terms that other pages related to the business, such as their Facebook page, also rank for?

To give a better idea of what the site looks like and how minimal it is, here’s a preview I generated with chatgpt. This screenshot isn’t an image of the exact site, all of the text on it has been changed slightly but this is more or less how the site looks. The real one is a different brand, tagline, address etc.

Essentially, it’s a brand name, phone number, a short sentence about the services provided and a tagline. Notice even the city and state are omitted from the address.

I originally wanted to make it even more minimal but I decided to build somewhat of a real site just in case any real customers happen to land on it. It is my cousin’s real business after all (and he doesn’t know about my little experiment and I want to remain in his good graces when he eventually finds out about this and I have to hand this site over to him lol)

The term I settled on targeting was “rock chip repair [city, state]” since none of those words (again not even the city and state) appear on the site and I knew from past link projects that I could quickly build some relevant links from auto service directories with this term.

So that’s what I set out to do. Well, not strictly auto industry directories, but also some local directories and (mostly) high authority general business directories.

For this test I wanted to stick with directories for a few reasons:

• People insist that directories have no value. Which is true to an extent, but that’s mostly because those people don’t know what they’re doing. If you’re spamming shitty directories, yeah, it’s likely not doing anything or even actively harming your site. But if you’re taking the time to find relevant niche directories they can be just as valuable as any other link, especially on moderated and trusted high authority sites.
• I wanted to test the effectiveness of directory links in isolation of other onsite variables. If there’s no content on the site, can these links do anything to improve positions for phrases that aren’t even alluded to on the page? You’d expect contextual links in a blog post or news story or whatever to be able to do this, but directory listings?
• I wanted to test their effectiveness in isolation of other types of links. How well can I move the needle with only directories?
• I can directly control many aspects of the link, including most importantly the anchor text but also the surrounding text and other aspects like which category I put the listing in. Some of them I can even control the URL.
• I can build more of them in a shorter amount of time. Honestly this is the main factor. If I could have quickly and reliably built enough backlinks with exact match anchor texts through manual outreach to “real sites” I would have but it doesn’t work that way.
• (Good) Directories are good.

I put together a doc of the links I built for this project and sorted them roughly in order of which links I personally think have the most value.

Funny enough I didn’t check to see which were nofollow until after I had finished sorting them based on which I felt had the most value, so it was amusing to me to see all of the nofollow links towards the top. Interestingly, two of the lowest DR ones are are ones that I personally feel are among the most valuable. These two in particular are: 1) a local community website that lists businesses and has an active facebook group with over 10,000 members, 2) an Idaho county newspaper’s business directory.

Following these, there are a couple high authority, well trafficked business directories, one of which I could control the anchor text on. Then a couple autoglass specific business directories and then some other various general business directories, many of which I was able to get exact match anchor text from.

When we start to hone in on the “best” links in this list, there’s one link that really seems to stand out from all of the others. They’re all good links in my opinion, or I wouldn’t have built them, but this one in particular hits a couple good marks.

1. exact match anchor text
2. high DR
3. not nofollow
4. (this doesn’t matter but some people think it does) the link appears in google for a site:URL search.

None of the other links check all of these boxes, although two other links hit points 1 and 3. I’m resisting the urge to turn this into a rant about how these metrics don’t really matter that much but, still, as SEOs we like to rank things so I guess there has to be one “best” link.

Is that link responsible for all of the performance boost we’re getting from this effort? I don’t know, but interestingly this is also the link that gets referenced when I ask chatgpt about rock chip repair in [city].

I started all of this at the beginning of May. I believe the first links I built for it were on May 6th.

By May 23rd or so we can see the site started appearing for rock chip repair for the first time and quickly rose to position one where it has bounced around a bit.

However, the most surprising part of this was seeing it appear in the AIO, which really has no reason to deliver the website instead of the Facebook page other than the fact that I built a bunch of links about rock chip repair.

Worth noting that with Google, the company’s Facebook page usually ranks somewhere between 1 and 3 for “rock chip repair [city]” and the website currently appears somewhere below the Facebook page the majority of the time for the organic results for most queries. The site was previously usually around 3-4 for the brand name before I built these new links but lately it’s been #1 every time I’ve checked. The facebook page also ranks for a lot of window related terms that the website doesn’t appear for (again the website doesn’t mention these anywhere and the facebook page does. Also it’s Facebook.)

At the time of this writing the site is sitting at 8 in the organic results in addition to the AIO.

However, on Bing the website is the only page associated with this business that ranks for this term, which is actually one of the more interesting takeaways for me. That means it’s appearing on bing for this term exclusively because of these backlinks/anchor texts. In my opinion.

Conclusions

So what’s the lesson here?

Honestly, I don’t know if there is one. We know links work. We know anchor text is important.

But I wanted to see how I could get things to move for a term that isn’t mentioned in the content anywhere.

I actually had no intention of testing the AIO specifically. In fact, the AIO only shows up for me for this term about 30% of the time and google does seem to be testing whether to return the fb profile or the website.

Though it does appear in the AIO often enough for Ahrefs to notice it at least.

How do I know this didn’t happen simply because the links were giving the site more authority and Google already correlates the brand for these terms? Or that the links simply allowed it to be crawled more often. Or anything else?

The truth is I don’t. Other than the fact that the site doesn’t really appear in the same way for other similar terms (it’s been bouncing around page 4 -6 for a couple of other similar but unrelated terms* that were mentioned frequently in the directory listing descriptions) and the timing of this starting to happen a week or two after the initial links were built. I’ve had the site live for about 6 months before doing this and it never appeared for these terms before.

*And here we can see that shortly after these links were built the site got a couple lonely impressions for these terms. My guess is the site has very little staying power due to being so barebones so it probably showed for one person and then never again. Yeah it’s almost no traffic (I would prefer if zero users ever actually land on the site in it’s current state honestly) but I’m sure the links have nothing to do with this:

Still, I think an honest linkbuilder will say “it could be a lot of things helping to drive this, but I know my work didn’t stop it from happening” and that goes for just about anything we do.

There are so many different factors at play especially regarding off-site SEO that it often doesn’t feel right to me to claim with 100% certainty that this did that, even if I know that if I want that to happen I should do this.

And I know the main critique of this already: that’s an insanely uncompetitive term, try doing the same in Los Angeles.

Yeah, no shit, in order to examine something in isolation you need to be able to isolate it.

Is this going to work in every vertical for every competitive term? Probably not. But then again, if you’re a good SEO you should be able to find some instances where you’ll be able to do something similar even in much more competitive arenas. Although, yeah, if you have a real brand and can build enough relevant links you probably can succeed at something like this.

But to me the point isn’t that we should make a strategy out of doing things like this, or that it’s a good use of time (because it probably isn’t in most cases) but instead I think a more important takeaway is that even in [current-year], with all of our constant algorithm updates and AI and everything else that seems to never stop changing, it’s still in no small part… backlinks and keywords. The same as it ever was.

In practice, it rarely works this way. Actually, the conversions on this style of linkbuilding aren’t typically any better than simply finding relevant pages to pitch your link to without the broken link strategy.

There’s a few reasons for this:

• Webpages tend to die of old age. This means that the page linking to them is likely also old and therefor not being updated any longer.
• Other linkbuilders have beaten you to it by now. If you’re in a competitive niche, the probability that someone else has found and outreached these targets is high. You’re digging through the literal garbage.
• These pitches are extremely easy to automate. If a site hasn’t received emails about the the particular link you’re reaching out to them about, it’s still likely they’ve received a pitch regarding a different dead link, and most of these emails are painfully the same.
• You don’t know the original intent behind that link. I mean, if it’s a piece of content, you can throw it into wayback machine and recreate it or even create a superior resource; if it’s a home page link, your service might genuinely be a better offering. These things will certainly improve your odds on a case by case basis but they don’t change the fact that links are created for a purpose and your site wasn’t part of that original purpose. And honestly if you’re able to sell your link to them under these circumstances, it’s likely they would have seen value in your link regardless of the broken link pitch.
• If the original link is now dead, what assurance do they have that yours wont eventually also die?
• They just don’t care.
• They don’t have edit privileges, whoever wrote it no longer works there, etc.
• They’ll simply say “thank you for letting me know, I’ve removed the broken link.” This is by far the most common, if you receive a reply at all. Sometimes they delete the page entirely (refer to the first bullet).

The list goes on.

So in addition to needing to “sell” your link in the same way you would any other link pitch, you also have to sell them on the mere idea that there’s any value in fixing that broken link. You’ve now created multiple barriers out of what once sounded like a brilliant strategy.

Most people who would replace a dead link with your link because of a cold email likely would have linked to you if you had simply asked them to anyways.

This doesn’t mean broken linkbuilding is a bad idea. Personally, I think even if something has a bad conversion rate it can still be worth doing. If your goal is to secure more links, then you should do more things to secure links.

However, the way most linkbuilders and nearly every guide on the topic tackle broken linkbuilding is flawed for all of the reasons I’ve outlined above and more.

I just took a break from writing this to read through all of the guides currently ranking on the first page for “broken link building” and all of them describe the exact approach of “find a broken link, offer yours as a replacement” and offer little more. Mostly they’re tutorials on how to find broken links, which maybe I should write about too but this is about the outreach. Brian Dean’s guide sort of goes into the first of my following tips but I can tell immediately that the example he gives of how to implement it will still fail the majority of the time - which is likely why his screenshot didn’t include a reply from the recipient.

Here are some things to try instead. These will all improve your broken link building efforts but they work even better when you can combine two or more of them at once:

Find a page on the same site to replace the dead link

In cases where a link is dead or broken but the website still exists, if you can find a replacement for that link that exists at the same website your suggestion becomes immediately helpful rather than simply using a dead resource to promote your own. It shows you’re not just automating* pitches, you’re putting in real effort to solve an issue for them. Here you can see this approach helped me land a great .edu link less than 2 hours after my email was sent.

*Even though you can totally still automate emails like this.

Let them know about multiple broken links

As in my previous screenshot, letting them know about multiple broken links can create a sense of urgency to fix them. It shows you’ve actually looked at the page and comes off much more helpful than simply asking someone to replace a singular broken link with yours. Also try letting them know about broken links on different pages of their site.

Find a broken link on a different page than the one you want a link on

If you’ve found a link opportunity it can be a good idea to search their site for any broken links you can also mention in your outreach. This also helps to make your outreach appear more personalized (even though you can easily automate this as well) by showing that you’re actually browsing their website. Another benefit to this one is it doesn’t have to be a broken external link. You can find broken internal links or other issues with their website that you can help them fix, or something like maybe the link to their Facebook profile in the footer doesn’t work. There are a lot of ways you can play this one and the best part is almost every website will have an issue you can report to them.

Find adjacent resources

This is also exemplified in my first screenshot. Instead of searching for resources that yours can be an exact replacement to, search for ones that have an intersecting relevance. For example, if you’re trying to build links to an article about home safety for senior citizens, you may search for a dead resource about disability accommodations, find a replacement for that dead disability accommodations website and then pitch your resource alongside it. You can search manually by Googling things like “disability accommodations guide inurl:resources”* but personally I like to let these opportunities find me. While you’re doing your regular link prospecting, keep an eye out for dead links that are adjacent topics to yours and you’re bound to discover great opportunities.

*I just Googled that exact phrase, checked for broken links on one of the first pages that stood out to me and found 4 great targets in about 20 seconds:

Offer someone else’s resource as a replacement

If you come across a dead link and aren’t able to find a suitable replacement for that resource at the same website, instead of offering yours as a direct replacement you could consider finding someone else’s to send. I like to do a quick Google search for whatever the title of the dead resource was and send them the best non-commercial site or one that isn’t from a direct competitor. Then after you’ve made that recommendation you can also include one of your own and suggest they add it as well.

Find dead resources from highly authoritative sources and fix them

Find dead links from places like non-profits or government agencies that pertain to your niche and then find resources at those same sites to fix them with. Simply take a site, in this case it was the CDC website, throw it into Ahrefs’ Best By Links and filter for 404s, then you can just search through the URLs for keywords pertaining to your link. Then it’s just a matter of finding a resource to replace it with. Start with the title of whatever the broken resource was and a site: search. If you can’t find a good replacement, you can also send them an archive.org cache of that page. I’m pretty sure in this particular example the guy would have added my link without the broken link pitch (again, most people who would replace a dead link with your link because of a cold email likely are likely to link to you regardless), but this is also an effective way of finding targets you might not find through other means - this one would have never been on my radar otherwise.

Mention the broken link in a followup

If you’ve already emailed someone your link request and haven’t received a reply, letting them know about a broken link on their website can be a great way to try to get the conversation going with them without sending them a bland followup email and it also shows that you’re a real person actively looking at their site. You can also easily schedule things like this as your followup email.

Mention the broken link after you’ve pitched your link

Instead of beginning the email by mentioning the broken link (like I’d guess nearly all other broken link emails do), mention the broken links at the end of the email. This will allow your actual pitch to take precedence in the email as well as demonstrates that you’re not trying to hide your request under a guise of helpfulness. This approach shows that you’re self aware in what you’re asking and also respectful of their time. Then, once you’ve made your pitch you can say “by the way, I noticed…”. I’d also guess that most broken linkbuilding email titles contain the words “broken link” in the subject. This email’s subject was literally just the name of the tool I was sharing with the recipient. I like this approach because it puts my resource out front and at the end gives them an extra reason to open up their website to make the edit.

Send a broken link to someone who has given you a link before

If someone has given you a link in the past and you have another link that you’d like them to add, letting them know of any broken links they have can be a great way to rekindle a conversation with them and an easy way to bring up an additional link you think they’d be interested in. If someone has been kind enough to link to you in the past, it can feel rude to reach back out to them saying essentially “can you add this one too?” without giving them anything in return.

Here’s on I rekindled from a year and half prior. (.edu link btw).

Redirects are better than 404s in a lot of cases

I like to find 301s for my “broken link building” and I find it to work much better than 404s in a lot of cases. Sometimes this is due to the age thing I mentioned early in the article (a link to a more recent blog post that now redirects to the site’s home page is often less ancient than a link to a site that is now completely dead) but mostly I think it’s because this approach is extremely overlooked. Most linkbuilders doing broken link building are looking mostly for 404s so this opens up a lot of opportunity. Throw any site into Ahrefs, go to Best By Links and filter for 3xx status codes. Look for ones that redirect from blog posts to the home page or even the main blog page. Here’s one site I found that has hundreds of links redirecting from various blog posts back to their home page.

Personally, because I have no intention to deceive or use these circumstances to upsell unnecessary services, it’s sometimes a bit of a challenge for me to answer these concerns in a way that feels like it would be satisfying to them. Because when clients read sensationalist takes about AI optimization, it feels a lot like a cop out to say “I’m already doing it” without making any huge flashy change from the way I’ve been operating even if that’s the honest truth.

If you’re a client and you read nearly anything about AEO/GEO/whatever these days, that’s probably going to be your takeaway. We need to do more. But what? Who even knows, just… something… more!

However, I started thinking about some ways to more tangibly illustrate my overall stance which is that most of this work ultimately boils down to crawlability (tech) and discoverability (I’m lumping content, brand, social, etc under this word. Further, when these things are done to their potential they all end up being to the aid of links anyways 🙂), and links. In short: make sure the bots can access your site and build authority.

When I was a teenager my grandpa started a home and auto glass repair business in a small Idaho town. I worked there for a couple summers. Because their shop is right in the middle of the town and there was really only one other place that offered similar service, they never built a website and never really needed one. Mostly they’ve relied on foot traffic and eventually a Facebook page. My cousin owns the shop now and still has never built a website for the business. They still don’t need one in my opinion.

As expected, they are the shop that chatgpt recommends. No surprise here, they’re the most established business in town at this point. Although there are more competitors now that the town has more than doubled in size since this business was established, this was still nice to see.

I followed up to see if chatgpt could find a website. I knew it wouldn’t, I just asked this because I knew I’d need it for this post.

Yep, no website. Cool.

These circumstances present a nice little opportunity to run a quick test. Most of the time the site comes first and the brand gets built along side it.

So I built two identical websites for this shop with two nearly identical domains. Two branded domains with a slight variation (one single character in the domain is different). On one I blocked all robots besides googlebot in robots.txt and Cloudflare’s AI crawl control settings and the other blocked google but allowed LLMs.

About a month later I started to build slightly different backlink profiles for each site. There was probably no need to wait but I wanted to see if either of these domains would be discovered on their own, which they weren’t. Initially I thought about building backlinks from entirely the same websites for each domain but decided against it because I want to do a few other experiments with these sites in the future.

For the site that I planned to let google index, I built a couple links from popular home repair (niche specific) sites that I could create business profiles on, just to see how quickly the site would get indexed from these links alone. For the other site I built them a couple dozen regular business directory links. I assumed Google would have no problem finding the site I wanted it to with a couple citations from the more “authoritative” and relevant sites and the LLMs would be more likely to discover the site I wanted them to with a larger amount of links. This ended up being true in this case pretty quickly. I did basically zero other “SEO”. Actually, other than doing the bare minimum to make sure the sites would be discovered, I’ve pretty much done the opposite of what nearly anyone would consider “best practices”.

The site essentially contains zero information about the business other than their name, logo, phone number and partial address (street name and number but no city or state mentioned anywhere on the site, I kept these details to a minimum just because again I plan to do some other tests with it eventually. However, this is relevant here because when we ask “what’s the best __ in [city]” they have no problem correlating the website with the business location at least partially because the city is listed in the directory links I’ve built). The site also doesn’t list any services other than what essentially says “we fix windows”. I wanted the site content to influence things as little as possible in order to get a more accurate idea of what these backlinks are actually doing here. This is the entire site:

Google indexed the site I wanted it to pretty quickly (within a few days of building those few links to it) and, no surprise, does not return the blocked site in a site: search (note: this does not mean google hasn’t indexed it).

It was also quickly ranking for its brand name. Nice, things are working more or less as expected:

A week or two later, I checked to see how things had changed in chatgpt:

Now if we click on the business name, it gives us additional information including a website. No surprise it’s the one I didn’t block LLMs on.

And a query from today (taken from a separate chatgpt account) as I write this post, just to see how things are still looking:

And from Claude?

So what do we learn from this?

Honestly? Literally nothing haha. Someone insinuated that LLMs aren’t able to use links in this way so that was the main inspiration behind this and then this pretty much all worked out exactly as I expected it to. I also needed to do it to lay the foundation for some future experiments. It was fairly surprising to me how quickly things moved though! Especially with these particular links which I think most people, including myself at times, would write off as ineffective.

I will say, however, that I’m pretty sure if I give it some more time and build more higher quality links to the one that the LLMs are blocked from eventually they will start recommending that website instead of the one they’re allowed to access. We’ll see what happens as I run through some other link experiments with these.