Morgan: Can you hear me okay?
Jake: Yeah, you’re good. Audio’s a little choppy but I can hear you.
Morgan: Sorry, I’m on hotel WiFi. I’m at this conference in Austin and the internet is—
—basically terrible.
Jake: [laughs] Classic. Which conference?
Morgan: MozCon. You here?
Jake: Nah, I’m in my office in Denver. Well, “office.” It’s my spare bedroom. My wife keeps threatening to reclaim it.
Morgan: [laughs] Fair. Okay so I’m just gonna jump right in because I only have like 30 minutes before my next session. You built a site with 100,000 pages?
Jake: Yeah. 107,000 to be exact. Travel comparison site. “Best hotels in [city],” “Things to do in [location],” that kind of thing.
Morgan: And you used GPT-4 to generate all of it?
Jake: Yep. Every single page. Well, not entirely true. I wrote the templates and the prompts, but yeah, GPT-4 did the heavy lifting.
Morgan: When was this?
Jake: Summer 2023. June through September. It was… [pause] …it was the fastest I’ve ever built anything in my life.
Morgan: Walk me through it. How do you even start a project like that?
Jake: Okay, so background first. I’m the Technical SEO Lead at this startup called TravelMatch. We’re basically a meta-search engine for travel bookings. Think Kayak but newer and nobody’s heard of us yet.
Morgan: Got it.
Jake: And in June 2023, my CEO— guy named Brandon— he’s obsessed with programmatic SEO. He keeps sending me articles about Zapier and Wise and how they scaled with programmatic content. And he’s like, “Why aren’t we doing this?”
Morgan: Uh oh.
Jake: [laughs] Right? So I start researching, and I realize we have this massive database of travel destinations. Like 12,000 cities, millions of hotels, all this structured data just sitting there. And I think, what if we could turn every city into a landing page?
Morgan: That’s pretty standard programmatic SEO though.
Jake: Totally. But here’s where it gets interesting. Instead of just creating basic “Best hotels in [city]” pages with the same template, I wanted to make each page unique. Real content, not just database fields filling in blanks.
Morgan: So you used GPT-4.
Jake: Exactly. I spent like two weeks in June building this whole system. Python scripts that would pull data from our database, send it to the GPT-4 API with really detailed prompts, and then automatically publish the generated content to WordPress.
Morgan: What did the prompts look like?
Jake: Super detailed. Like, I’d give it the city name, population data, top attractions, average hotel prices, weather patterns, local events— everything. And then I’d ask it to write a 1,500-word guide that felt like a local wrote it.
Morgan: Did it work?
Jake: Dude, it worked so well it scared me. The content was good. Like, actually good. It had personality, it had specific recommendations, it had this conversational tone that didn’t feel robotic at all.
Morgan: So what happened next?
Jake: I ran a test in early July. Generated 100 pages for different cities, published them, and just… watched. Within a week, we were ranking for a bunch of long-tail keywords. Within two weeks, we were getting traffic. Real traffic, not bots.
Morgan: What kind of numbers?
Jake: That first test? Maybe 500 visitors a day across all 100 pages. Which doesn’t sound like a lot, but our entire site was only getting like 2,000 visitors a day at that point. So it was a 25% increase from 100 pages.
Morgan: And then you scaled it up.
Jake: [laughs] Oh, I went crazy with it. I was like, if 100 pages gets us 500 visitors, what does 100,000 pages get us? So I just… ran the script. Over and over. I’d generate 5,000 pages a day for like three weeks.
Morgan: Jesus. Did anyone try to stop you?
Jake: Brandon loved it. He was watching the analytics like a hawk, and every day our traffic was going up. By mid-August, we were at like 35,000 visitors a day. By September, we hit 50,000.
Morgan: That’s insane growth.
Jake: It was incredible. And we weren’t just getting traffic— we were getting conversions. People were clicking through to booking sites, we were earning affiliate commissions. In August alone, we made like $87,000 in revenue. September was even better, like $143,000.
Morgan: Wait, so in two months you made over $200K?
Jake: Yeah. From basically nothing. Brandon was talking about raising our Series A based on these numbers. I was getting job offers from other companies. It felt like I’d cracked the code.
Morgan: When did it fall apart?
Jake: October 11th, 2023.
Morgan: What happened?
Jake: I woke up, made coffee, opened my laptop to check our morning analytics, and our traffic was down 40%.
Morgan: Overnight?
Jake: Overnight. I’m staring at Google Analytics thinking there’s a bug. So I check Search Console, and… [pause] …60,000 of our pages had been deindexed.
Morgan: Oh shit.
Jake: Yeah. I refreshed the page like five times thinking it was a mistake. But no, 60,000 pages just gone. Not ranking lower. Gone from Google entirely.
Morgan: Did you get a notification? A manual action?
Jake: Nothing. No email, no Search Console warning, nothing. They just silently deindexed more than half our site.
Morgan: What did you do?
Jake: [exhales] I didn’t tell anyone at first. I just started frantically Googling. “Google deindexed pages,” “why did Google remove my pages,” all this stuff. And I’m finding forum posts from other people who did programmatic SEO at scale and got hammered.
Morgan: Did you figure out what triggered it?
Jake: Not immediately. I spent the entire day trying to find patterns. Which pages got deindexed versus which ones stayed. And I started noticing something.
Morgan: What?
Jake: The pages that survived were the ones for major cities. New York, LA, Chicago, Miami. The pages that got nuked were for smaller cities. Like “Best hotels in Pawnee, Oklahoma” or “Things to do in Cedar Rapids, Iowa.”
Morgan: Why do you think that mattered?
Jake: I think Google’s algorithm looked at those pages and went, “There’s no way anyone actually wrote 500 detailed guides about tiny cities in Kansas.” It was too obvious that they were auto-generated.
Morgan: But the content was good, right? You said it had personality.
Jake: It had artificial personality. Like, the content looked good if you read one page. But if you read ten pages, you started seeing patterns. Same sentence structures, same types of recommendations, same voice. It was like… [pause] …it was like talking to the same person with different masks on.
Morgan: What did Brandon say when you told him?
Jake: [laughs bitterly] Oh, that conversation was fun. I waited until end of day because I wanted to have some kind of plan first. And I walk into his office and I’m like, “We have a problem.”
Morgan: How’d he react?
Jake: He freaked out. Not like, yelling at me. But he pulled up the analytics and just watched the numbers for like two minutes without saying anything. And then he goes, “Can you fix this?”
Morgan: Could you?
Jake: I didn’t know. I told him I’d try, but I honestly had no idea if those pages would ever come back.
Morgan: What did you try first?
Jake: I deleted the worst offenders. Like, pages for towns with populations under 5,000. Places where we obviously had no real expertise. That was about 20,000 pages.
Morgan: That must have hurt.
Jake: Dude, I felt sick doing it. Each page I deleted represented potential traffic and revenue. But I figured if Google thought those pages were spam, keeping them up would just make things worse.
Morgan: Did it work?
Jake: Not right away. The deindexed pages stayed deindexed. But at least we stopped bleeding. Like, no more pages disappeared after I cleaned up the obvious junk.
Morgan: What about the pages that were still indexed? Did they stay?
Jake: Yeah, and that was weird. We still had like 47,000 pages indexed and they kept performing. Traffic didn’t go back to where it was, but we stabilized around 30,000 visitors a day.
Morgan: So you were still making money.
Jake: Some. October revenue was down to like $95,000. November was $88,000. Not terrible, but Brandon was pissed because we’d gone from growth mode to damage control.
Morgan: Did you try to get the deindexed pages back?
Jake: I tried everything. I submitted them to Search Console for recrawling. I added more internal links to them. I even went back and manually edited like 500 of them to make them more unique.
Morgan: Did any of that work?
Jake: Some came back. Maybe 10,000 pages got re-indexed over the next few months. But like 50,000 are still gone. Just… permanently disappeared.
Morgan: How much revenue do you think you lost?
Jake: [pause] If we’d kept growing at the rate we were going, we’d have hit maybe $2 million in revenue by the end of the year. Instead, we finished at around $800K. So I cost the company like $1.2 million in potential revenue.
Morgan: Did Brandon blame you?
Jake: Not directly. But he definitely cooled on me. Like, I used to be the golden boy. The guy who grew traffic 2000% in three months. And then I became the guy who broke the site.
Morgan: Are you still at TravelMatch?
Jake: Yeah, but my role’s different now. I’m not leading growth stuff anymore. I’m mostly doing technical audits and maintenance. It’s fine, but… [trails off]
Morgan: It’s not what you want to be doing.
Jake: No. [pause] I mean, I learned a ton from this whole thing. But it also kind of broke my confidence, you know? Like, I used to think I understood SEO. Now I’m not sure I understand anything.
Morgan: Do you think what you did was wrong?
Jake: [long pause] I don’t know. I mean, was the content helpful? Yes. Was it unique? Sort of. Was it spam? [pause] Maybe. I don’t know where the line is anymore.
Morgan: What would you do differently?
Jake: Slower. Way slower. Instead of generating 100,000 pages in a month, I’d generate 100 pages a month and actually monitor how Google responds. Build trust gradually instead of trying to speedrun it.
Morgan: Do you still use AI for content?
Jake: Yeah, but completely differently. Now I use it for outlines and research. Then humans write the actual content. It’s way slower and more expensive, but at least I can sleep at night.
Morgan: Do you think programmatic SEO is dead?
Jake: [pause] No, but I think the gold rush phase is over. Like, you can’t just spin up 100,000 pages overnight and expect Google to reward you anymore. You have to actually provide value at scale, which is way harder.
Morgan: What does Brandon think about all this?
Jake: He thinks we were early to a trend that Google shut down. Like, we were onto something, but we pushed it too hard too fast. He might be right.
Morgan: Are you looking for other jobs?
Jake: [laughs] Maybe. I’ve had recruiters reach out. But honestly, I’m not sure what I’d tell them in interviews. “I grew a site to 50,000 visitors a day and then lost half of it”?
Morgan: That’s still impressive growth, even if you lost some of it.
Jake: Yeah, but it feels hollow. Like, anyone can grow traffic if they’re willing to break rules. The hard part is sustainable growth that doesn’t blow up in your face.
Morgan: [phone buzzing] Shit, sorry, I have to run to my next session. This was really helpful though.
Jake: No worries. Hey, you gonna write about this?
Morgan: Probably, yeah. Is that okay?
Jake: Sure. Just… [pause] …maybe don’t mention TravelMatch by name? Brandon wouldn’t love that.
Morgan: Of course. I’ll keep it vague.
Jake: Thanks. Good luck at the conference.
Morgan: Thanks. And hey, for what it’s worth? I don’t think you did anything wrong. I think you were just playing the game everyone was playing.
Jake: [pause] Yeah. Maybe. I appreciate that.
Morgan: Alright, I gotta run. Talk soon?
Jake: Yeah, anytime. Bye Morgan.
Morgan: Bye Jake.
[end of part 1]
Part 2: Three Months Later
[This is a follow-up call conducted in January 2024. Morgan and Jake reconnect to discuss what’s changed since their first conversation.]
Morgan: Hey! Can you hear me?
Jake: Yeah, you’re good. Way clearer than last time.
Morgan: [laughs] Yeah, I’m in my actual apartment now, not a hotel. So, it’s been what, three months since we talked?
Jake: Almost exactly. How was the rest of the conference?
Morgan: Good! Actually, a lot of people were talking about programmatic SEO. Like, it was a whole track. Made me think of you.
Jake: [laughs] I bet half those talks were cautionary tales at this point.
Morgan: Pretty much. But I wanted to check in because I’ve been following your site and… are you guys growing again?
Jake: [pause] Yeah. We are, actually. That’s what’s so weird about all this.
Morgan: Wait, really? Tell me what happened.
Jake: Okay, so after we talked in October, I was in full panic mode for like six weeks. Just trying to understand what went wrong, trying to salvage what we could. And then in December, something shifted.
Morgan: What shifted?
Jake: Our traffic started growing again. Not from the deindexed pages— those are still gone. But the pages that survived? They started ranking better. Like, way better.
Morgan: Why?
Jake: I think Google rewarded us for cleaning up the junk. Like, once we deleted all those questionable small-city pages, our overall site quality improved. The 47,000 pages that remained were all for legitimate destinations where we actually had good data.
Morgan: So you’re back to where you were before?
Jake: Not quite, but close. We’re at about 42,000 visitors a day now. Revenue in December was $131,000. We’re on track to do $150K this month.
Morgan: That’s amazing! Is Brandon happy again?
Jake: [laughs] He’s happier. He’s still mad we’re not at $200K a month, but at least we’re growing. And he’s talking about funding again, which is a good sign.
Morgan: Are you creating new pages?
Jake: Yeah, but super carefully. I’m adding maybe 200 pages a month now. All major destinations, all with heavy human oversight. I have two freelance writers who review every page before it goes live.
Morgan: That must be expensive.
Jake: It is. We’re spending like $8,000 a month on content review. But it’s worth it. I’m not going through that October nightmare again.
Morgan: Have any of the deindexed pages come back?
Jake: A few hundred here and there. But the bulk of them? Still gone. I’ve accepted they’re probably never coming back.
Morgan: How do you feel about that?
Jake: [pause] Honestly? Relief. Like, those pages were always a ticking time bomb. Now I know exactly what I have, and it’s all solid. No more wondering if Google’s going to nuke us again.
Morgan: Did you change anything else about your approach?
Jake: Yeah, a few things. One, I’m way more conservative about keywords now. I only target terms where we can actually provide value. Two, I’m obsessive about content quality. Every page has to pass what I call the “would a human write this?” test. And three, I’m diversifying traffic sources.
Morgan: What do you mean by diversifying?
Jake: We’re investing in social media, email marketing, even some paid ads. I don’t want to be 90% dependent on Google anymore. That was stupid.
Morgan: That makes sense.
Jake: Yeah. And honestly, it’s made my job more interesting. I was getting bored just generating pages at scale. Now I’m thinking about holistic marketing strategy, and it’s more fulfilling.
Morgan: What about your relationship with Brandon? Did that recover?
Jake: It’s getting there. He brought me back into strategy meetings last month, which is a good sign. I think he realized that what I built was still valuable, even if I screwed up the execution.
Morgan: Do you feel like you screwed up the execution?
Jake: [long pause] Yes and no. I think I pushed too hard too fast. But I also think the whole industry was doing that. Like, everyone was racing to scale with AI, and I was just one of many people who got burned.
Morgan: True. I’ve talked to like six other people with similar stories.
Jake: Really?
Morgan: Yeah. E-commerce sites, local business directories, recipe blogs. Everyone tried to scale with AI in 2023, and a lot of them got hit.
Jake: That makes me feel less alone, honestly. I was starting to think I was uniquely stupid.
Morgan: [laughs] No, you were just early to a common mistake.
Jake: [laughs] I’ll put that on my resume. “Early adopter of common mistakes.”
Morgan: [laughs] Hey, can I ask you something else?
Jake: Sure.
Morgan: If you could go back to June 2023 and do it all over again, what would you change?
Jake: [pause] I’d still build the system. The programmatic approach was sound. But I’d cap it at like 10,000 pages instead of 100,000. I’d focus on quality over quantity from day one. And I’d diversify traffic sources immediately instead of putting all my eggs in the Google basket.
Morgan: Would you still use GPT-4?
Jake: Yeah, but differently. I’d use it to generate first drafts, then have humans heavily edit them. Add real expertise, real local knowledge, real personality. Make it so that even if someone read 100 pages, they wouldn’t notice a pattern.
Morgan: That’s basically what you’re doing now.
Jake: Exactly. I just wish I’d figured that out before I lost 60,000 pages.
Morgan: [pause] Do you think programmatic SEO has a future?
Jake: Absolutely. But it’s not about scale anymore. It’s about programmatic processes that enable human quality at scale. Does that make sense?
Morgan: I think so. Like, using systems to make human work more efficient rather than replacing humans entirely?
Jake: Exactly. The dream of “press a button and generate 100,000 perfect pages” is dead. But the reality of “use smart systems to help humans create great content faster” is very much alive.
Morgan: That’s actually a pretty optimistic take.
Jake: [laughs] Yeah, well, I’ve had three months to process my trauma. I’m in the acceptance stage now.
Morgan: [laughs] Fair. Alright, I should let you go. But hey, thanks for the update. I’m glad things are turning around.
Jake: Thanks for checking in. And hey, if you publish that interview, send me the link?
Morgan: Definitely. Take care, Jake.
Jake: You too, Morgan.
[end of part 2]
Key Lessons Learned
“The dream of ‘press a button and generate 100,000 perfect pages’ is dead. But the reality of ‘use smart systems to help humans create great content faster’ is very much alive.”
1. Speed Kills in Programmatic SEO
Jake’s biggest mistake wasn’t using AI—it was the pace. Generating 107,000 pages in three months triggered Google’s spam detection. The same strategy executed over 12-18 months might have succeeded.
2. Patterns Are Poison
Even high-quality AI content reveals itself at scale. When Google (or users) can detect repetitive sentence structures, similar formatting, or predictable content patterns across hundreds of pages, the algorithmic fingerprint becomes obvious.
“The content looked good if you read one page. But if you read ten pages, you started seeing patterns… It was like talking to the same person with different masks on.”
3. Not All Pages Are Created Equal
The pages that survived Google’s deindexing purge were for major cities where TravelMatch could legitimately claim expertise. The pages for tiny towns with populations under 5,000 were obviously low-value targets created solely for SEO.
4. Recovery Requires Sacrifice
Jake had to delete 20,000 pages—representing significant potential revenue—to signal to Google that he was serious about quality. Sometimes you have to amputate the infected limb to save the patient.
5. Dependency Is Dangerous
Being 90% dependent on Google organic traffic meant that a single algorithmic action could destroy the business. Jake’s post-crisis strategy of diversifying across social, email, and paid channels is essential for any SEO-dependent business.
6. The “Would a Human Write This?” Test
Jake’s new quality threshold is simple: would a human actually sit down and write this specific page? If the answer is “probably not,” it shouldn’t exist—regardless of how good the AI-generated content looks.
7. Sustainable Beats Explosive
Jake went from generating 5,000 pages per day to 200 pages per month. His growth slowed dramatically, but it became sustainable. Revenue went from $143K (peak) → $88K (crash) → $131K (recovery). Slow and steady won the race.
“Anyone can grow traffic if they’re willing to break rules. The hard part is sustainable growth that doesn’t blow up in your face.”
8. AI Is a Draft Tool, Not a Publishing Tool
Jake’s evolved approach treats AI-generated content as a first draft that requires significant human editing, local expertise, and personality injection. This hybrid model preserves speed while maintaining quality.
9. The Deindexing Shadow
Three months after cleanup, 50,000 of Jake’s deindexed pages remain gone—likely permanently. Google’s memory is long, and once pages are flagged as spam, they rarely recover. Prevention is everything.
10. System Design > Content Volume
The future of programmatic SEO isn’t about generating more pages faster—it’s about building systems that enable humans to create better content more efficiently. Process automation, not content automation.
About Jake Torres
Jake Torres is a Technical SEO Lead with 9 years of experience in search marketing, specializing in large-scale website optimization and programmatic content strategies. A graduate of the University of Colorado Boulder with a degree in Computer Science, Jake began his career as a web developer before transitioning into SEO in 2015.
His early career was marked by steady success—he helped a Denver-based e-commerce company increase organic traffic by 300% through technical optimization and earned a reputation as a problem-solver who could navigate complex technical challenges. By 2022, he’d moved into leadership roles, managing SEO strategy for high-growth startups.
Jake joined TravelMatch, a travel meta-search startup, in early 2023 as their Technical SEO Lead. What followed was one of the most dramatic growth-and-crash cycles in recent SEO history: a programmatic content experiment that generated \$400K in revenue over four months, followed by Google deindexing 60% of the site and a \$1.2 million revenue shortfall.
“I used to think I understood SEO. Now I’m not sure I understand anything.”
The experience fundamentally changed Jake’s approach to SEO. Where he once chased rapid growth through automation and scale, he now advocates for sustainable, quality-focused strategies that balance efficiency with human expertise. His current methodology—using AI for drafts but requiring extensive human review—has become a model for other technical SEOs navigating the AI content landscape.
Since the October 2023 deindexing crisis, Jake has successfully stabilized TravelMatch’s organic traffic at 42,000 daily visitors and rebuilt revenue to pre-crisis levels. His willingness to discuss his mistakes openly has made him a sought-after voice in conversations about the responsible use of AI in content creation.
“The future isn’t about programmatic scale—it’s about programmatic processes that enable human quality at scale.”
Outside of work, Jake is an avid rock climber and homebrewer. He lives in Denver with his wife, Sarah (a UX designer), and their rescue dog, Algorithm (yes, really). He occasionally speaks at SEO conferences about technical SEO and the lessons learned from his programmatic content experiment—though he admits these talks are “basically therapy sessions in front of an audience.
Jake remains at TravelMatch, where his role has evolved from pure growth tactics to holistic marketing strategy. He’s currently working on a book about sustainable SEO practices in the AI era, tentatively titled “Scale Without Breaking Shit.”
Editor’s Note
This interview was conducted in two parts: the initial conversation took place in October 2023 during MozCon in Austin, Texas, and the follow-up occurred via video call in January 2024. The interview has been edited for length and clarity, though we’ve preserved Jake’s authentic voice and the natural flow of conversation.
Jake requested that his company name be changed to protect their business interests. “TravelMatch” is a pseudonym, though all other details—including revenue figures, traffic numbers, and the timeline of events—are accurate.
This multi-part format allowed us to capture not just the crisis moment, but also the recovery process—a reminder that SEO disasters aren’t always permanent, and that learning often happens in the aftermath.
