Log 001 Fixing Google Ads after they simply... stopped working - Sorilbran Stone (1)
|

Intent Beats Keywords: The Real Fix for Underperforming AdS

Originally published May 30, 2025 to The Builder’s Log – Log 001 on LinkedIn

Builder’s Log Entry 001: Fixing the Ads

Initiative: Google Ads Rebuild & Realignment
Approx Start Date: February 2025
Starting State: 5-fig monthly investment. Zero sales-qualified leads

Quick Context

This experiment started as a paid media rebuild.
It became a lesson in AI visibility, intent alignment, and machine-readable trust.

I inherited Google Ads the same month I became a one-person marketing team. No one handed me a baton or even gave me a heads-up. It was more like walking into a room, hearing something beeping in the corner, and realizing, “Oh. This is mine now.”

I’d never managed paid search before. Never logged into Google Ads. I didn’t even know what a “Responsive Search Ad” was. So I asked AI to teach me. That was February. This log was written in May.

I wanna walk you through what’s happened since and why it’s about way more than just ads.

Caveat: Technology can do this – all day long. But the insights will help you work with the technology you hand this work off to.

Phase 1: Learn the Lingo

The first thing I realized? Our agency partners were speaking fluent Google but that wasn’t translating into results. Everything they shared looked like success… until I dug into the lead quality and conversion paths.

They were optimizing for click-through rates. We needed real buyers. Actual revenue. And there was a massive disconnect between how they defined success and what success actually looked like for us. We were misaligned.

So I used Maverick (my AI assistant) to teach me the basics: ad structure, CTR benchmarks, CPC expectations, industry norms. I needed to see through the polish and get to the substance.

Phase 2: Strategic Misfires

Back in November, I’d created segmented signup pages based on audience intent. By January, those pages were finalized. But when I inherited the ads in February… none of them were running to those pages.

Not one.

When I checked GA and HubSpot, I couldn’t find evidence that most of those pages had ever been used.

And the pages that were being used? They were mismatched and underperforming. Badly. CTRs below benchmarks. CPCs climbing. Zero demo requests. Just traffic… wandering.

So I paused. Not the ads – but my instinct to crash out. Instead, I started tracing the funnel backward.

Phase 3: Intent Over Keywords

Our setup at the time was built around keywords. Logical, sure – but too shallow. People were clicking, but they weren’t converting. And worse, the landing pages weren’t helping them make a decision. They were just… there.

So I rebuilt the strategy around intent clusters. That way, I can be sure that the headline visitors see in ads align with the language they see on the landing page.

I used Grain to analyze 24 calls tied to 15 lost deals. And in almost every one, the same pattern showed up. These weren’t bad leads. They were just under-informed. The marketing hadn’t done its job.

There was a clear gap between what the prospect wanted to believe about us and what they actually understood by the time they got to Sales. That’s on marketing. They should have arrived better equipped through sales enablement content, redirects to FAQs, smarter landing pages, whatever it took.

So I had Maverick distill those pain points into a matrix of objections and misunderstandings we could anchor our ad copy to. The goal was simple: educate early, so Sales doesn’t have to play both therapist and translator in the same call. By the time a prospect makes it to Sales, they should be aligned, with only logistics left to work out.

That’s the job now. Make sure the ads do the pre-work.

 

Phase 4: Landing Pages That Don’t Convert

We tested three things:

  1. Sending traffic to signup pages → ❌ No lift.
  2. Sending traffic to landing pages → ⚠️ Slight engagement bump
  3. Sending traffic to case study pages → 🟡 Promising, but incomplete

The one consistent pattern? Every SQL we got hopped from the landing page to a case study before they converted from a button none of our landing pages was tied to – the button in the nav menu.

That told me two things:

  • Trust > Traffic
  • Nobody wants to hop into a car with you if they don’t even know if your brakes work (okay, I learned that one before this, but you get my point)

So I started adding case studies to our landing pages. But then I hit another wall.

Phase 5: Case Studies Were Invisible

ChatGPT wasn’t citing us. Perplexity was confused. Maverick started reading invisible fields from our website that didn’t match what I saw.

Turns out, two things were amiss:

Thing 1: The theme-native way we formatted our case study blocks was largely invisible to AI. Not exactly invisible, but it wasn’t the glowing review we thought it would be. I’ve since added code to properly tag individual case study blocks with meta data: this is a case study block about a campaign in this vertical that solved this problem and ran on these platforms.

Thing 2: The way our templates were built included hidden rows from other clients. That meant our Mandarin orange campaign case study also had backend data from a gay dating app and an industrial lighting company.

You cannot make this stuff up.

Worse: Because the fields were hidden from users but not machines, AI environments were getting a scrambled signal – and either hallucinating, or ghosting us entirely.

Phase 6: AI-Readable ≠ Human-Friendly

I tried using JSON-LD to structure the data. And that worked… kind of. Machines liked it. But users – not so much. Between mobile use and shorter attention spans, long, text-heavy case studies with zero interaction or scannability? That’s a no.

So now I’m designing case studies for humans and machines:

  • Semantically structured HTML blocks
  • Clean visual formatting
  • Modularity for remixing across pages
  • Clear storytelling arcs optimized for AI + conversion

Not as fancy and flashy as our previous templates, but clean, responsive, and machine-readable – that’s the new standard of good content.

The first few case studies took days, trying to land on a design that didn’t totally unravel our brand aesthetic and obscure our brand voice. Believe it or not, our brand voice and aesthetic bring in top-tier talent and serve as powerful differentiators in a crowded market.

Phase 7: One Problem Leads to All Problems

This started as a Google Ads fix. Now it’s a full-blown ecosystem cleanse.

Because AI has become a major source of our traffic. And that traffic stays longer and engages deeper than search or social. And because everything from our CRO to SEO to Paid to Content to Comms is now connected by the same visibility layer.

If it’s broken in one place, it’s probably broken in six.

Results – June 3, 2025

  • Cut annual ad spend in half, pipeline is actively rebounding
  • Reduced # of ads by 85%

Generic Ads (Q2)

  • CTR: 4.06% (up from 3.66%)
  • CR: 11.34% (up from 2.14%)
  • CPA reduced by 84%

Branded Ads (Q2):

  • CTR: 9.65% (up from 7.25%)
  • CR: 7.29% (up from 1.67%)
  • CPA reduced by 60%

Dig Deeper