A developer at a German bank asks an assistant which payments providers support a particular scheme with a German licence. It names three. If your English documentation is thorough and your German regulatory pages are thin, you may be described accurately as a technology and inaccurately as unavailable.
That is the specific exposure for a Frankfurt technology company in this layer. It is not primarily a visibility problem. It is a description problem: what a model can say about you is bounded by what you have published, and most fintech companies have published a great deal about how their product works and very little about the conditions under which it can be bought.
The classic results page did not expose this, because a reader who arrived on your documentation would eventually find the licensing page. A generated answer does not browse. It answers from what it can attribute, and an unattributed condition is an absent condition.
Ten results become three descriptions
A conventional listing offers ten links and leaves the judgement to a person. A generated answer performs the judgement and produces two or three names with a sentence each. The sentence is the part worth thinking about — it is a description of your company written by something that read your site.
What ends up in that sentence is not the best-optimised material but the most attributable. A selection needs facts that can be tied to the question: which scheme, under which licence, in which jurisdiction, with what settlement arrangement. Documentation describing an elegant API supplies none of them, however good it is.
What a technology company routinely omits
Facts that exist internally and appear nowhere public
All four are known to compliance and to enterprise sales. None of them are usually on a public page in German.
- Which entity holds which authorisation, and where. Named regulator, named permission, named jurisdiction. This is in every due-diligence response you have ever sent and almost never on the website.
- What happens to funds and data. Where client money sits, where data is processed, what applies in an insolvency. These questions stop deals and are answered privately, one prospect at a time.
- Who can actually buy. Institution types, minimum volumes, jurisdictions served and — more usefully — those not served. Stating the exclusions is what makes the inclusions credible.
- How long adoption takes. Not integration time in engineering days but elapsed time from first contact to production, including procurement and review. Everyone knows the number; nobody publishes it.
The common property of these four is that publishing them costs almost nothing in effort and something real in comfort. Stating who cannot buy, or how long adoption actually takes, forecloses conversations that sales would rather keep open. That is the genuine reason they are missing — not oversight, but reluctance — and it is worth naming, because the trade is a favourable one: fewer conversations, a higher proportion of them viable.
What can be examined without a counter
Absent a tally, what is left is method: put the same questions repeatedly, archive what comes back, and read the pattern. Six surfaces in the Semalt workspace are built for exactly that, presenting the field the way a model characterises it instead of the way a results page orders it.
- Market Circle placement. Leading, middle or niche tier. It reports distance from the names reached for first, which for a company with strong engineering and weak commercial pages is frequently worse than expected.
- The model's own account of your domain. Positioning, an estimate of traffic, openings it identifies. For a fintech this is the single most useful screen, because it shows what your published material actually communicates.
- Questions sorted by intent. Phrased questions rather than term lists, separated by whether somebody is orienting, comparing or preparing to buy.
- Existing pages flagged as leverage. Published pages where more content or better internal linking would pay. On documentation platforms these tend to be pages nobody in marketing has looked at.
- The topics rivals have finished and you have not started. Areas another domain treats end to end while nothing comparable exists on yours. Here that is almost always the regulatory and operational side rather than the technical one.
- One consolidated figure for AI-assisted search. A single value across the layer, and precisely the item the remaining two caveats govern.
The second surface is the one to run first. Where the model's account of your company reads as a description of a developer tool while your revenue comes from regulated institutions, that gap is not a modelling failure. It is an accurate summary of a site that documents an API thoroughly and its licensing conditions barely at all.
What the consolidated figure cannot bear
The figure stays useful for as long as its provenance stays visible. It comes from trials: questions asked, answers recorded, appearances counted. That is a sample rather than a census, and the distinction matters more here than in most industries.
In a company that sells to regulated institutions, publishing an unqualified figure carries a second cost beyond inaccuracy. Prospects in this segment audit their suppliers, and a number in a deck that cannot be reconstructed on request damages more than the number. Stating the three limits alongside the value is not caution for its own sake; it is the difference between a figure that survives scrutiny and one that invites it.
One organisational note before the practical part. Everything described here sits at a boundary that most technology companies have not drawn: the regulatory and commercial facts belong to compliance and to enterprise sales, while the website belongs to marketing. In practice that means the material exists in one place and the publishing capability in another, with no established route between them. Companies that establish that route once — a standing arrangement to convert approved private answers into public pages — solve the problem permanently. Companies that treat each page as a separate negotiation publish two pages a year, which is not enough to matter.
The reason to say this plainly is that the analysis will point at the same gap every quarter until the routing exists. Reading the model's account of your domain four times and noting the same absence four times is not progress; it is an expensive way of confirming something you learned in the first quarter. The workspace will keep reporting it, and the report is only useful if somebody can act on it.
English documentation, German decision
A pattern worth naming, because it catches Frankfurt companies specifically: the question that names your company and the question that excludes you are frequently asked in different languages by different people at the same institution.
"Which providers support this scheme?"
Asked by engineering. Answered from documentation, where you are strong, and you appear.
- Technical attribution is easy
- Rarely the deciding conversation
"Who is licensed for this in Germany?"
Asked by compliance or procurement. Answered from regulatory material, where you may be absent entirely.
- Requires named permissions in German
- Frequently the deciding conversation
Appearing in the first answer and not the second is a common and expensive position, because it produces exactly the outcome that looks like a technical loss and is not: the engineering team liked the product, the approval never happened, and nobody involved could articulate why. The remedy is a small number of German pages containing named permissions and named conditions — which is an editing task drawing on documents that already exist.
The licensing page in German
Names the entity, the regulator and the permission, in the language the compliance function reads.
- Matches a jurisdiction question directly
- Written once, revised annually
The English trust page
Describes certifications and a security posture in general terms, with no named permission or jurisdiction.
- Matches no specific question
- Reads as marketing to a reviewer
Attributable sentences
| Typical phrasing | Attributable replacement | What it enables |
|---|---|---|
| "regulated in Europe" | named regulator, named permission, named entity | matches a jurisdiction question |
| "enterprise-ready" | institution types served, with those not served | matches a buyer-type question |
| "fast integration" | elapsed weeks from contract to production | supplies a comparable figure |
| "secure by design" | where data is processed and under which regime | answers a due-diligence question |
| "trusted by leading institutions" | a named client type and a year | ties a claim to something checkable |
None of these replacements is a marketing improvement in the conventional sense — several read as less impressive. They are, however, the only versions that can be used when something has to choose between three suppliers, and they answer the questions a German institution asks before it will start a procurement process at all.
Why the four facts stay unpublished
Each objection is reasonable, and each has a narrower answer than it first appears.
- "Publishing our exclusions costs us leads." It costs the leads that would have been declined after two calls. The volume drops and the proportion that closes rises — which is the trade most enterprise teams say they want when asked directly.
- "Our licensing position is complicated." It is, and you already explain it in writing to every serious prospect. The complexity is an argument for publishing carefully, not for publishing nothing.
- "Adoption timelines vary too much to state." A range with the conditions attached is both accurate and usable. What cannot be stated is a single number with no conditions, and nobody was asking for that.
- "Competitors would see it." They already have it, from your prospects. The people who do not have it are the buyers deciding whether to start a process at all.
Working through these four once, with sales and compliance in the same room, is what unblocks the material. Attempted separately it stalls indefinitely, because each function has a partial objection and neither can resolve the other's. The page recommendations in the workspace help in that meeting, because they name specific existing pages rather than proposing an abstract programme of work.
A rhythm that fits a shipping team
Because the underlying evidence here is weaker than the familiar metrics, a weekly rhythm is out of the question. Three-monthly fits, and it lines up conveniently with the interval at which regulatory text has to be reread in any case.
| When | What happens | Who does it |
|---|---|---|
| Week 1 | Collect twenty questions verbatim from the sales inbox and due-diligence responses | whoever handles enterprise enquiries |
| Week 2 | Check which of them has a public page in German | content |
| Week 3 | Read the model's account of the domain and the competitor coverage | content, thirty minutes |
| Weeks 4–8 | Publish two or three pages built from existing internal answers | content plus one compliance review |
| Quarter end | Compare placement and the consolidated figure with the previous quarter | nobody senior, deliberately |
The first row is where the material comes from, and it is close to free. Due-diligence questionnaires are answered anyway, at length, by people who know the answers. Turning ten of those answers into public pages requires a review rather than an author — and the review is quicker than usual because the text has already been approved once for exactly this purpose. That analysis sits beside the conventional figures in My SEO, so a publication and any subsequent movement appear on one timeline.
What this produces
Before committing effort, it is worth being clear about the output. Not a metric for an investor update. A site that states the conditions under which the product can be bought — and therefore performs better in both layers, while shortening the sales cycle for reasons unrelated to search.
That last effect is worth stating plainly because it usually exceeds the search benefit. Every question answered publicly is a question not answered again in a call, and in enterprise sales those calls are the expensive part. A company that publishes its licensing position, its exclusions and its realistic adoption timeline spends less senior time on prospects that were never going to close. The technical prerequisites for any of it are covered under technical SEO, the editorial split between English and German material in our content strategy, and which German terms are worth pursuing at all in keyword research.
See how your domain is described in the dashboard
Questions from product, content and compliance
Can we measure how often we are cited?
Not in the sense of taking a reading. No provider has access to a tally. The workable substitute is putting a fixed set of questions on a schedule, archiving the replies and reasoning a direction out of them. That reasoning does the job it is meant to do and stays a series of trials; asked to behave like a traffic statistic it will fail, and in front of an auditor that failure is expensive.
Our documentation is strong. Why are we described as a developer tool?
Because that is what the published material describes. A site with thorough API documentation and a single paragraph about licensing communicates a developer tool, accurately. The description is not wrong; it is incomplete in exactly the way the site is incomplete. Publishing the regulatory and commercial conditions changes the description because it changes what there is to read.
Should we publish which institutions we cannot serve?
Yes, and it is the single highest-return item on the list. Stating exclusions makes the inclusions credible, filters out enquiries that would have consumed senior time before failing, and gives a selection something specific to match against. Sales teams object to it consistently and change position within a quarter, once the composition of incoming enquiries shifts.
Does any of this require new compliance review?
Less than expected, because most of the material has been reviewed already. Text drawn from completed due-diligence responses has been through approval once, for a more demanding audience than a website. What usually remains is a check that the wording is still current and that nothing entity-specific has been carried across — a shorter review than a new page from scratch, and a materially different conversation to have with a reviewer.
How often should the consolidated figure be reviewed?
Once per quarter, and by the person doing the work rather than by a manager. Inspect it more frequently and you are watching the trials wobble. What the number means becomes clear only next to the familiar figures: a shifting placement against unchanged positions and traffic is worth a closer look, while both moving in step nearly always shares one explanation.