Reality Production

AI produces the derivative.We produce the primary source.

We produce first-hand evidence from your site, translate it into machine-readable form and measure where you are named across four systems.

Nothing published without your approvalNo ranking guaranteesOnly measured results get reported

The boundary first: this is not a production service

If you are looking for a corporate film, this is the wrong address; many good companies make those. What is sold here is a production line: deriving a shot list from search demand, producing evidence on site, translating it into machine-readable form and measuring the result. The film is only one of that line's outputs.

What you are buying

Three steps. Each stands on its own; the next is proposed only when the previous one justifies it.

Visibility audit

Entry, fixed scope

We record what the four systems say about your company today, what they do not know, and from which source they learned what is wrong. Output: a zero-line report and an inventory of false statements. If no production is needed, we say that too.

Production sprint

Core work

A shot list derived from search demand, the production day, the asset package, structured data and distribution. We do not sell shooting days; we sell a system. The delivery is not a single file but a linked body of assets.

Monitoring and growth

Ongoing

Regular measurement under the identical protocol, publication of new assets, follow-up on corrections. AI answers fluctuate; a single measurement is no measurement.

Scope and budget are set per project; we do not publish a price list. For comparison: a modest trade fair budget lives a few days, while the evidence archive stays on your own domain and compounds.

How it works

Without all three legs the line does not work. The missing leg shows which half of the job was skipped.

The camera documents

The image produced in the studio or on site is a record of fact, not of claim: the running machine, the measured tolerance, the completed test. The shot list follows search demand and real buyer questions.

The system translates

The image alone is not enough. Transcript, page text, ImageObject and VideoObject markup and IPTC metadata tell the machine who documented what, where and when.

Measurement proves

The zero line is measured before publication and again afterwards under the identical protocol. Across ChatGPT, Gemini, Claude and DeepSeek we record what is said, which source is cited and what is stated incorrectly.

Why now

The ground under this approach is not an opinion. The findings below come from public sources and are not TYS measurements.

Adding statistics, expert quotes and source citations raised visibility in generative answers by 30 to 40 percent. The largest gains went to small, low-authority sites.
Princeton and Georgia Tech, GEO study, 2024
The signal most strongly correlated with AI visibility was YouTube mentions; classic backlinks stayed far weaker. Correlation is not causation, but the direction is clear.
Ahrefs, analysis of 75,000 brands, 2025
Google announced updates that push original content and first-hand perspectives forward: the Highly Cited label and preferred sources carried into AI answers.
Google, 2026 announcements

What one production day yields

A properly planned day does not produce a single film but a linked tree of assets.

Video and short clips

With chapter markers, transcript and description text. In AI citations, formats that show and explain carry high weight.

Measurement and test records

Tolerance, duration, durability, capacity. A shoot that produces numbers is worth more than one that produces pretty pictures.

Photo corpus

Angle, detail, scale reference, material texture, environment of use. Every frame is identified with IPTC metadata.

Technical page and article

Published as HTML, with short machine-extractable answers. What sits in a PDF catalogue enters no answer chain.

Structured data layer

VideoObject, ImageObject, Service and FAQ markup, matching the visible page content exactly.

Distribution and measurement record

Publication, indexing and crawler access. That the machine read it is shown through server logs.

What we commit to and what we do not

Our commitments

  • Measurement before and after under the identical protocol, with a report
  • Indexability and machine accessibility of every asset produced
  • The false-statement inventory and publication of corrective content
  • No single frame published without your approval
  • Clear labelling of whether a movement is noise or change

What we do not promise

  • Ranking guarantees. Nobody can give them; whoever does is mistaken.
  • Guaranteed mention in AI answers. Answers are probabilistic.
  • Results in thirty days. Changes usually surface after weeks.
  • Reaching into the model itself. What moves is the retrieval layer.
  • Traffic targets in numbers. This is about a few high-value queries.

Frequently asked questions

We already have a corporate film. What is different here?

A production company delivers the film and the work ends. Here the film is one line item; the actual delivery is the shot list built from search demand, machine-readable metadata, structured data, distribution and measurement. Measuring and reporting what the video achieved is part of the engagement.

We do not want to show our process, we have trade secrets.

Instead of process proof we film result proof: the measuring room, the test rig, certificates and calibration, material traceability, macro shots that hide geometry. Capability can also be demonstrated on a purpose-designed benchmark part. Details on the industry page.

How is AI visibility measured?

With a frozen question set, several repetitions per system and measurement waves across multiple days. A single measurement is no measurement: research from the University of St. Gallen found that repeating the same query on the same day yields cited sources that overlap only 32 to 43 percent. A noise band is therefore declared.

Is it not cheaper to produce content with AI?

Cheaper yes, and worthless for exactly that reason. Models do not cite text that repeats what they already know. New knowledge comes only from reality: the test performed, the measurement taken, the real footage captured. The same material also serves the sales conversation and the trade fair stand.

How long until results are visible?

Content changes usually surface after days to weeks; first citations take several weeks, robust visibility takes months. The zero line is measured on day one so that progress becomes provable.

Which industries do you work with?

The first focus is manufacturing exporters with visually rich production processes. The approach works wherever there is provable reality.

The starting point is always measurement

First we look at what the four systems say about your company today. If the measurement is good, we say so; we do not sell production nobody needs.

Request a visibility audit