Source First, Then Echo

When an AI estimated our prices

In August, we asked an AI what Aivis-OS costs. At that time, our prices were not yet on the website. The answer came anyway: 6,000 to 12,000 euros one-time, plus 150 to 300 euros per month. Where the numbers came from remained unclear. They did not originate from us.

We drew a simple conclusion and published our prices. The bigger question remained: If gaps are filled by others anyway – shouldn’t you go external first, into technical articles, directories, and mentions?

External sources count – a lot

The question is valid. According to Muck Rack (May 2026), 84 percent of AI citations originate from third-party editorial sources; according to AirOps (2026), 85 percent of brand mentions come from external domains. In our own sample for a travel market provider, several answers relied on comparison portals; the provider’s website was cited by only one of the systems. Ignoring external sources would therefore be wrong.

But it depends on what they propagate. And that is where the problem begins.

What happens when your own source offers two versions

In one audit, a website cited two different values for the same metric in different places. The AI took neither. It took a third one, from a PDF document.

If your domain offers two versions of your operational scope, a further technical article does not decide which one applies. It might simply propagate both. And even if buyers never visit your website, your content is read and summarized. That is why your domain matters: it is the one source whose content you determine yourself.

External anchors also require verification

Even the references intended to make a company unique via structured data can be off-target. In one project, we had linked a data product to a Wikidata entry of the same name – which described a broadband product from a completely different industry. The name matched, the entity did not. In another, a company’s Wikidata entry listed an incorrect headquarters. Since then, our rule is: external entries are anchors to uniquely identify a company. We do not consume values from them without verification.

Fill the gaps first

Back to your own source. Gaps are not filled on suspicion, but scenario by scenario – before external sources are aligned. An example: An industrial client is looking for a maintenance partner for three plants, two of them in Austria and Switzerland. Once the operational scope is clarified, their next question might be: What happens if a plant fails at night? What response time applies to which plant? A clarified operational scope does not answer these questions. The missing answers are obtained, approved by those responsible, and added.

Then the echo

Only then are directories and profiles aligned with the approved version, where agreed. Partners and platforms can carry the information further, even to places where you have no direct access. Good SEO and GEO work (optimization for AI responses) remains useful. It just needs a source it can build on.