#015 - Your Website Has a Second Audience Now - Human in the Loop Newsletter

September 21 2026

I’ve spent more than 20 years helping businesses explain themselves on the web.

For most of that time, the goal was straightforward: help people understand the business and help search engines find it. AI has made that work interesting all over again.

AI systems now read your website and the wider web to decide what your business does, who it serves, whether it can be trusted, and when it should be recommended.

This became impossible to ignore while Marius and I were doing the same audit work for several clients at Agentsy. We kept finding established businesses with strong reputations, years of useful content, and plenty of real proof being reduced to incomplete or outdated answers.

The existing tools could tell us whether the business appeared and give us a rating. They rarely explained why the result looked that way or what the business should do about it.

So we did the builder thing and turned our manual process into a product. You can try it at Agentsy.report.

Building it forced us to decide what software could collect and analyze for us, and where human judgment is still needed.

Why we left out the score

When we started building the tool, we considered giving every business an AI visibility rating.

Almost every other tool does it. A single number is easy to understand, easy to compare, and easy to track over time.

But the deeper we got into the client work, the less useful that number felt.

Two businesses can receive the same rating for completely different reasons. One may have weak authority and very little useful content. Another may have years of strong evidence trapped in PDFs, scattered across old pages, or described more clearly by third-party websites than by its own.

The same problem appears when tracking progress. A headline metric can change because the questions, providers, or test conditions changed. It can also hide meaningful progress happening underneath it.

We saw both patterns in our client work.

One established business we work with received poor AI-readiness results despite having stronger proof, certifications, operational history, and content than its main competitor. It didn’t need a new identity or strategy. It needed to make its existing substance easier for AI systems to retrieve and cite.

Another client showed encouraging progress after a round of improvements. The company’s own website was being retrieved and used as a source more often, even though the broader mention rate had not caught up yet (it has now!).

A single rating would have obscured the most useful part of both audits: understanding why the result looked that way and deciding what the business should do next.

So we left the numeric rating out. Agentsy.report shows the underlying measurements, evidence, and findings instead.

What we automated, and what we kept human

Leaving out the score made the product harder, but more interesting, to build. We needed to collect and organize enough evidence to support a deeper conclusion.

So we went deep.

Agentsy.report reads the business’s website, tests the questions buyers might ask across multiple AI systems, preserves the answers and sources, looks at competitors, and examines the site’s SEO, performance, and technical readiness. It also generates the initial findings that help us identify possible patterns.

That creates a lot of material.

We built an internal workbench where Marius and I can inspect the findings, trace important claims back to their sources, compare answers across AI systems, and add our own analysis.

Internally, we need that depth. The business owner shouldn’t have to work through hundreds of raw findings to understand what matters.

So the same evidence has to work in two different ways. It needs to be detailed and traceable enough for our review, then clear and focused enough to become a useful client report.

That is where we draw the line between automation and human judgment.

Agentsy.report collects the evidence and generates the initial insights. Marius and I then help the business owners work through those findings: confirming what is accurate, identifying what matters commercially, and deciding which changes deserve attention first.

Put it to the test

Agentsy.report is live now.

The free scan gives you an immediate view of how AI currently interprets your company. Try it with your business and compare the findings with what you know to be true.

The interesting part is often the gap between the two. AI may rely on old positioning, overlook an important service, miss the evidence behind your claims, or recommend competitors in situations where your business should be considered.

The Audit goes much deeper, working to explain those gaps rather than simply identifying them. It tests how your business appears throughout the buyer’s decision, traces influential answers back to their sources, and examines your competitors, website, SEO, performance, and technical readiness.

Marius and I then help you make sense of that evidence, confirm which findings are accurate, and identify the changes most likely to improve how your business is understood and recommended.

You can run the free scan here:

Or start with the full Audit:

The larger takeaway is that AI now needs to be treated as another audience for your website and your business. What it understands will increasingly influence which companies buyers discover, trust, and choose.

If you want to go deeper

I’ve also written two guides that explore parts of this problem in more detail.

How AI Engines Work explains how the various AI systems use model knowledge, live web retrieval, and outside sources to construct their answers, along with what that means for your website.

Authority and Proof looks at why making your website accessible to AI is only the beginning. Being understood and trusted also depends on clear authorship, firsthand experience, original evidence, and credible sources.

This is only the beginning

Agentsy.report focuses on how AI understands and recommends a business. But that is only one part of the shift.

AI systems are also beginning to interact directly with websites: researching products, completing forms, booking appointments, and taking actions on someone’s behalf.

That creates a new set of questions about how businesses should structure their websites and services for an audience that may not be human at all.