Cyrus Shepard disavowed every link to his own website. Twice. Nothing happened. That is the first thing Patrick at Elite Strategies wanted me to ask him about, and it turned into 51 minutes on what actually moved sites after the Helpful Content Update: brand searches, branded anchors, author photos the size of a billboard, and pulling internal links out instead of putting them in.
You probably know Cyrus from Moz, where he led SEO on and off for a decade. He runs Zyppy SEO now, out of Astoria, Oregon (home of the Goonies, no local clients, global work), and spends his time running experiments and publishing the data when he can. This is his second time on the show. The first, with Mark A Preston hosting, was about his years as a Google quality rater, and you can find that one at unscriptedseo.com/cyrus-shepard. This one is about what to change on your site this week.
🎧 Listen to this episode
The clip above is the 53-second version of his “70,000 articles with that exact title” point. The full conversation is audio only, so listen first, then read on for the parts you’ll want to quote.
Listen on Castos: https://the-unscripted-seo-interview-podcast.castos.com/episodes/cyrus-shepard-on-brand-signals-disavow-tests-and-de-optimizing-sites
Watch the clip: https://www.youtube.com/watch?v=5ouh3rbY4RE
Disavowing 3,219 domains moved nothing, three times
Here’s the sequence. In 2013, at Moz, Cyrus disavowed every link to his site he could find, about 1,500 domains, and waited. Nothing. In early 2024 he did it again on zyppy.com: every link out of Search Console, Ahrefs and Semrush combined into a file of roughly 3,000 root domains. Two months later, “nothing happens. anything, my traffic went up a little bit. I didn’t publish any new content.” He published that result and four theories came back at him: the tool does nothing, Google ignored a file that obviously disavowed everything, two months isn’t long enough, or the tool only fires in specific situations like a negative attack or a manual penalty.
So he ran it a third time and never published this one until now. A third of his links, deliberately including good ones, left in for about four months from last summer. Then he forgot about it. Someone asked a week and a half before we recorded, he checked, and the file was still in there. “You have disavowed 3,219 domains, took the file out and nothing has ever happened.”
“Maybe the disavow tool does something, but at this point, we can’t prove it. It’s all voodoo at this point.”
I gave him the counter-case. Michael McDougald had a competitor standing up 30 spun domains a month, sludging hot links with broken and inappropriate anchors at everybody in the niche. An intern found the new 30 each month, they went in a disavow file, and we recorded a positive movement. Our theory was that it was the anchor text diluting relevance, a secondary signal, not a direct hit. Cyrus’s read is that you use the tool when you’re penalized or under attack, but “the dream of SEOs that we can use the disavow tool to really positively influence our rankings, we want to use it as a Swiss knife, that’s really difficult.”
The things only SEOs do are the things Google flipped on
A year of HCU recovery work gave Cyrus one lens for “people-focused content”: what does a normal person do that you and I don’t? Your grandma types a post into WordPress and hits publish. You tweak the title tag for click-through, stack schema on it, add 30 internal links from the rest of the site, and (my contribution) fuss over the H2s. The sites doing that hardest fell hardest.
That hurt, because his own 23-million-internal-link study from before HCU showed anchor variety as “the strongest signal I’ve ever seen from the data.” After the update, “sites with lots and lots of internal anchor text variations are now the ones at the bottom of the pile.” His words for it: “it’s almost like Google went through and like, what are SEOs doing to manipulate our results? And then they just did the opposite, you know?”
The travel example makes it concrete. Search “things to do in New York City” in quotes and Google has something like 70,000 articles with that exact title. In December 2024 you’d have won that with keyword research, Surfer, subheadings and links, and you’d have been competing against every other site that did the same thing. What rose instead was a sleepy forum:
“How fortunate was Rick Steves that he just happened to have this sleepy travel forum, plugging along for 10, 15 years, and then Google decides they like people-focused content, and boom, Rick Steves is, you know, king of travel.”
Brand is not hand-wavy anymore, and you don’t need to be IBM
Tom Capper, his old Moz colleague, found that sites hit by HCU had a much lower share of branded search than sites that weren’t, and the pattern held through the March and August core updates. That lines up with what Mark Williams-Cook told me in the jam sandwich episode about the distribution of branded versus unbranded clicks. Cyrus goes a step further: people searching for you is one signal, but Google can also read how many people are talking about you, whether you’re in the knowledge graph, and how much of your anchor text is your brand. Those are things an SEO can move.
He’s blunt about how the advice has landed so far. “Telling a small publisher to build a brand is like telling a homeless person, just buy a house.” The fix he’s selling isn’t becoming a household name: “you don’t need to be IBM. You just need to tweak it a little bit. You just need to get the spigot flowing and get some of those brand searches coming in, get some of those web links and mentions.”
In practice that flips where the digital PR budget goes. The old pitch for a mattress site was a keyword-rich anchor, “best mattress 2025.” The new pitch is an article about the mattress experts at that company, what they think, why they’re experts, with quotes from the CEO and links to the people. “It’s no longer about the article and the exact match keywords. It’s about you and building the brand.”
His own cautionary tale: he named his newsletter “SEO Tips” because that’s what people search for. Anybody could write that newsletter, and as I pointed out, about 25 people do. “I should have called it, you know, the Zyppy SEO tips newsletter, put Zyppy SEO in there somehow. And it’s a small thing, but over time, those things make a difference.” One travel client had gone so far into reader-focus that he didn’t want a logo or navigation on the page. Getting a branded logo onto every page “was a huge struggle but we did it along with a bunch of other things”, and that site is one of the ones recovering. I wrote up this thread of the conversation, with two other guests’ takes, in The Evolution of SEO: From Keywords to Brand Authority.
Five things to change on your site this week
I asked him for the execution list, the thing you do differently the week after you hear this. Here’s his answer, in his order:
- Check whether any Google update in the last two years hit you. If none did, “keep doing what you’re doing because Google hasn’t applied a classifier on you and you’re doing well.”
- Measure your branded share of search against competitors. Moz Brand Authority gives you about 10 free searches a month, or do it yourself in Search Console. “You don’t want to be below the average. You want to be above the average. If you’re below average, you have some work to do.”
- De-optimize. “Probably one of our most effective strategies of recovering sites that have been hit by [HCU] and other updates is de-optimizing and removing internal links, making our title tags much more boring, making sure our titles match our H1s, things like that”. Mega menus, related-article blocks, the “12 most popular articles” at the end of every post: “we just wipe them out.” Owners panic about engagement, and “hardly anything ever happens”. The sites that went hardest on removal recovered best.
- Put the people into the content. His example is NapLab, a mattress review site that breaks every rule (over-optimized, too many internal links) and kept winning updates, because they jump on the mattress, shoot 50 photos, measure thickness and publish the data. “They’re doing things that don’t scale.” His Excel-formulas client now has videos of himself on every post in goofy poses. “It’s about you and building the brand.”
- Shift the PR spend from keyword anchors to articles about your people. Same outreach, same budget, different ask.
My version of the same thing is a mirror. Whatever your business is on this side of the glass, your job as an SEO is to make sure the reflection exists on the internet, and that is far easier if you do something real first. I launched a white-label trash cleanup service for SEO companies’ clients; Olaf led 20 volunteers in Sarasota off event pages I posted to Eventbrite, the county site and Reddit. Real thing, real photos, real mentions, no anchor text negotiated.
20% branded anchor text versus 2%: the number from the winners and losers
I ended by asking what surprised him in the last month, and he pulled up a file. His Moz-era studies were huge correlation runs across tens of thousands of domains with the outliers thrown out. Now he does what he calls “really bad data analysis”: take the top 1% of sites that have never lost a Google update since 2023, take the bottom 1% that have lost every one, and compare only those. He warns you up front that the method exaggerates correlations. Then:
“Sites that have never lost a Google update had on average 20% branded anchor text from different domains. The bottom 1% had on average 2%. And it was correlated ridiculously high, above 50%, which is not even real in the SEO world.”
He is using that to steer link building and says the results are positive, then immediately undercuts himself: “Or it could just be we’re building links and it works. So I don’t know.” That is the honest version, and it matches what we measured across our own network this year, where cross-site anchor diversity moves and anchor wording on its own doesn’t. Note the qualifier in his sentence, too: branded anchors from different domains. Twenty branded links from one site is not what he counted.
His close, when I asked what SEOs should take away: “Be real. Be authentic.” And then, borrowing Wil Reynolds’ line, “do real shit.”
Where to find Cyrus
He’s mostly on LinkedIn and Bluesky and will answer on X if you reach out. He also broke his no-plugs rule for one thing. He gets more leads than he can handle and has nowhere good to send them, because “marketing directories are just broken”: Clutch and DesignRush are “all pay to play”, and the Semrush and HubSpot partner lists require you to subscribe first. So he’s launching a human-powered, community-rated directory called Zyppy List and inviting consultants and agencies to apply.
- Website: Zyppy SEO
- Zyppy List: the launch announcement
- LinkedIn: Cyrus Shepard
- Bluesky: @zyppy.com
- His first episode, on Google quality rating: SEO from the eyes of a Google quality rater
Listen: https://the-unscripted-seo-interview-podcast.castos.com/episodes/cyrus-shepard-on-brand-signals-disavow-tests-and-de-optimizing-sites · Watch the clip: https://www.youtube.com/watch?v=5ouh3rbY4RE · Subscribe to the Unscripted SEO Interview Podcast wherever you get your podcasts so the next experiment lands in your feed before the write-up does.
