We Asked the Robots About Ourselves. Then We Fixed It.

There’s a particular kind of quiet that settles over a room when you realize you’ve been giving advice you haven’t taken yourself.

For months, I’ve been telling nonprofit leaders and purpose-driven business owners the same thing, “Your next customer, donor, or volunteer probably isn’t scrolling through ten blue links on Google anymore. They’re asking ChatGPT. Or Perplexity. Or reading the AI summary at the top of the search results. They’re getting three or four names back. And those are the names they’ll remember.”

Which raised an awkward question. “What do the robots say about Relish?” (Time to bust out the mirror.)

The Chasm of Mediocrity

We knew that our visibility online was dipping. Not only was the space getting more and more fragmented with the introduction and increase in use of AI-search, we hadn’t invested the time and energy in our own Generative Engine Optimization (GEO) to help boost our presence. (Cobbler’s kids and all.) To top it off, we were busy building a new offer, the AI Visibility Snapshot, based on the GEO work we have been doing for a few clients, and it didn’t feel right to sell it to a wider audience before running the system on ourselves. So we did. Same process our clients get: eight questions a real buyer would ask, run fresh on ChatGPT, Perplexity, Gemini, and Google’s AI Overviews – every answer scored the same way.

The good news came first. When you ask about Relish Studio by name, the robots are downright complimentary. ChatGPT and Perplexity called us “well-regarded,” quoted our 4.9-star Clutch rating, and praised our work with nonprofits. (We’ll chalk that up in the “win” column.)

Then came the questions that actually matter – the ones a buyer asks before they know we exist. “Best web design agency in Denver.” “Who should I call if my website isn’t bringing in customers?” “How do I choose a good web design agency in Colorado?”

Across 19 of those answers, Relish showed up exactly zero times.

Zero. Not ranked low. Not mentioned in passing. Absent.

And the kicker: when Google’s AI answered “Is Relish Studio any good?”, it helpfully mixed in employee reviews from a completely different company with a nearly identical name. A prospect doing their homework would likely have not noticed the mix-up.

Those results stung a little. (More than a little, TBH.)

What We Did About It

Here’s the thing about AI answers: they aren’t random. The robots recommend the organizations they can clearly understand – who you are, what you do, who you serve, and where – and the ones other trustworthy sources vouch for. So we did exactly what we’d tell a client to do. Over the past couple of months, we worked through a roadmap of on- and off-site optimization adjustments that included:

• Getting crystal clear on positioning. Relish works with purpose-driven organizations, nonprofits first. We said so. Explicitly. Everywhere.
• Fixing what was broken. Old pages, stale facts, and loose ends the robots were tripping over. That crusty, out-of-date offer from 2015 that no one had clicked on for 6 years but was still on our site? Gone.
• Building real founder pages. So “Who is Stu Swineford?” has a better answer than a LinkedIn snippet.
• Rewriting our service pages and case studies. Specific outcomes, specific clients, specific places.
• Adding structured data. The behind-the-scenes code that tells search engines and AI tools, in their own language, exactly what we are.
• Cleaning up our directory listings. Including an old office address (or two *cough*) that was still floating around years after we’d left.
• Giving the Audience Engagement Cycle a proper home. Our framework deserved a page that explains it well – and claims it as ours.

None of this is magic. It’s the unglamorous work of making sure your website and your listings tell one clear, consistent story. Customers seeking your services love this. Turns out… it’s what the robots want too.

Then We Asked Again

On October 26, we reran the exact questions from our first tracking check on September 27. Same wording, fresh sessions, three runs per question on each platform. Here’s what moved:

Across the six questions a buyer asks before they know our name, on ChatGPT, Perplexity, and Gemini, we went from winning 3 of 18 to 6 of 18. (A “win” means the robot recommended us in at least two of three fresh runs.) Perplexity now calls us “the clearest nonprofit-focused match” for Colorado nonprofit web design. Perplexity also credits us, in every run, for the Audience Engagement Cycle – four activities and all: Attract, Connect, Bond, and Inspire. (That last one made me especially happy. When someone asks a robot what the Audience Engagement Cycle is, the answer now points back to us.)

What Didn’t Move (Yet)

I’d be lying if I told you we’d won the internet. We haven’t.

• “B Corp web design agency Colorado” still goes to two certified B Corp agencies. All three engines read the question as “an agency that is a certified B Corp,” not “an agency that works with B Corps.” We went 0 for 9. Fair enough. We may need to rekindle our B Corp certification efforts.
• “Nonprofit marketing strategy consultant Colorado” still goes to other firms. (Gemini named us once in three tries. We’ll take it.)
• Denver is the head-scratcher. We rank #2 on Google for “Denver nonprofit web design agency,” yet only one of nine AI answers named us. Ranking on Google and getting recommended by AI are not the same thing.
• Broad questions like “best web design agency in Denver” are still a long climb to the top.

And a word of honesty about the numbers: AI answers shift from one run to the next. That’s why our September and late-October checks ran every question three times on each platform, and we only count a win when we show up in at least two of three. A quick one-run check on October 21 had us at #1 for Denver on Perplexity. Five days later, we went 0 for 3. That’s weather, not climate. We can’t point to a single change and say “that’s the one that did it.” But the direction we’re moving is clear, and it’s the right one to improve how we get found by an increasing number of people online.

What This Means for You

If you take one thing from our little experiment, let it be this: you can’t fix what you can’t see. Your Google Analytics won’t tell you when ChatGPT recommends your competitor. Your SEO report won’t either. The robot just quietly hands your customer to someone else, and nobody sends you an email about it.

Wondering how your organization shows up in AI search? A few lessons we’d pass along as you start to explore the results:

• Use the language your audience actually uses. Not your name – their problem. That’s where the gaps hide.
• Be specific about who you serve and where. “We help everyone” gives the robots nothing to work with.
• Hunt down your name twins and stale facts. Old addresses, old names, and similar-sounding companies all muddy the water.
• Claim your ideas. If you have a signature framework or program, give it a real home on your site. Otherwise, someone else gets the credit. (On one of our own runs, Gemini confidently invented new stages for our framework. Claiming your ideas is ongoing work.)

In Audience Engagement Cycle terms, this is Attract-phase work in a new arena. The robots are becoming the front door for a lot of your future supporters. Make sure they can find the doorbell.

Want to Know What the Robots Say About You?

That’s exactly why we built the AI Visibility Snapshot. We ask ChatGPT, Gemini, Perplexity, and Google’s AI Overviews the eight questions your customers (or donors) are already asking, then hand you a one-page scorecard showing who the robots recommend and what to fix first. Every Snapshot includes a walkthrough call with us.

It’s $99 for nonprofits and $199 for businesses, and if you’d like help with the fixes, your investment counts toward any Relish project you start within 30 days – an AEC Assessment, GEO fixes, even a new website.

Get My Snapshot

Source for results: Relish Studio AI visibility tracking, September 27, October 21, and October 26, 2026 (ChatGPT, Perplexity, Gemini, and Google AI Overviews; three runs per question on September 27 and October 26). AI answers vary from run to run.