For those of us who started with SEO back in the late ’90s, getting a priority listing on the early search engines was easy. That changed fast once Google entered the picture. Complexity went up, and so did the stakes: we had to figure out what worked, what didn’t, and what got us penalized. We could read what others claimed, but the only reliable way to know was to test it ourselves. That “Always Be Testing” (ABT) mentality started early, and it’s shaped every year we’ve spent in SEO since.
The early days: color-matched keywords and other bad ideas
One of the earliest ranking “hacks” we tried was hiding keywords by setting the text color to match the background. Sneaky, sure, but plenty of people swore it worked. We tested it ourselves, and for a while, it did. Then our test site got penalized and dropped from search entirely. That was an early lesson and thankfully one we didn’t add to client websites. But that lesson cemented ABT as our default posture rather than a one-off precaution. In our industry there’s a lot of noise out there so ever since then, we’ve tried to keep it to a minimum and of course ABT.
What ABT actually taught us
The real payoff of two decades of testing wasn’t any single tactic. It was learning to tell the difference between activities that consistently moved the needle and activities that were either risky, unproven, or just temporary loopholes. That distinction is where we’ve focused our energy ever since, rather than chasing whatever trick was making the rounds that month.
Same mentality, new front: AI search
ABT has carried us through 20-plus years of optimizing for Google, adapting as user behavior changed. The last couple of years have brought the biggest shift yet. Google is still central, but ChatGPT, Claude, Perplexity, and other AI tools have become a real part of how people search and research. That’s pushed us to apply the same ABT discipline to a new question: how do we help clients improve their visibility when prospects are researching through AI instead of, or alongside, traditional search? That shift is still moving, and the ground hasn’t settled.
The new hype cycle: AI visibility scores and unproven metrics
Just like the early SEO days, a new wave of self-declared experts has shown up with strong opinions, new tools, and confident “here’s what you need to do” playbooks. And just like before, a lot of it doesn’t hold up under testing. AI visibility scores, prompt tracking, AI analytics: what are these actually measuring? Do brand mentions inside an LLM’s output actually correlate with business results? And which prompts are you even tracking, when real prospects rarely phrase their questions the same way twice?
Why we still test everything ourselves
Trust matters more than ever here. We rely on a small group of people we know are rigorous testers, and sometimes their word is enough. Other times, it’s exactly what pushes us to go test something ourselves. We’re not extending blind trust to the new crop of AI-visibility experts, and we’re not trusting LLM output at face value unless we’ve fed it the right context first. Twenty-five years in, our mentality hasn’t changed: always be testing.
