Free check
For German machinery and plant engineering companies that export

Your best customers know you.The next ones ask AI first.

When a plant manager in Ohio or Ho Chi Minh City specifies a new line, he asks a chat window — in English — and gets three manufacturer names. We measure which three they are in your category, whether you are among them, what is said about you, and whether it is true.

Show me what AI says about us

Freeresults in 72 hoursa two-minute formno call unless you ask for one

How we measure
Exhibit 01 · sample measurement, placeholder data
Question

“Which manufacturers of blister packaging lines do you recommend for pharma?”

Answer

For pharmaceutical blister packaging, the manufacturers most often named are Supplier A, Supplier B und Supplier C. These suppliers cover lines from blister forming through to serialisation …

Page 1 / 3
Finding

Your company: not named

1–3 Competitors named, in the order the answer gave them.

Act I
The silent loss

The buyer you never meet.

Your long-standing customers are not going to leave you over an AI answer. The risk sits somewhere else.

It sits with the project engineer specifying a new line right now, who has never heard of your company. He would once have asked an agent, a colleague, or walked a trade fair. Today he types a question into a chat window and gets three names.

Those three names decide who even receives an enquiry.

If you are not among them there is no lost tender, no rejection, no phone call. There is only an enquiry that never arrives — and it appears in none of the numbers you track.

Act II
Correctness

Described wrongly costs more than not being named.

Not being named costs you an enquiry you never see. Being described wrongly costs you the ones you already have.

Four things we check in every answer about you: whether the performance figure quoted is current or comes from a discontinued series. Whether approvals — CE, ATEX, pressure equipment directive — are assigned correctly, not at all, or to the wrong product. Whether the application or format range named is yours or a competitor’s. Whether your name is being confused with a similar-sounding one.

Fact-check · sample extract, placeholder data
4 of 6 claims
Claim made by the AILikely sourceStatus
“Throughput up to 400 blisters per minute.”Product catalogue 2019, series discontinuedincorrect
“Lines carry CE marking.”Company page /quality, as of 2026correct
“Experience with EU FMD serialisation.”no source foundunsupported
“Format range up to 120 mm draw depth.”A competitor’s data sheet, confusion of namesincorrect

Your buyer puts that answer next to your data sheet. If the two do not match, it is not the model that looks unreliable. It is you.

And you cannot correct it by calling someone. There is no editorial desk to file a correction with. The answer changes only when the sources it is assembled from change. Which ones those are in your case is in the report.

Act III
Language

The answer that decides your next enquiry is assembled in English.

Your German pages are good. That is not the problem.

The problem is that a model researches predominantly in English-language sources even when the question is asked in German — and that your buyer in Ohio, Pune or Monterrey asks in English anyway. He gets an answer built from texts nobody at your company has ever worked on. Your English site is often a 2019 translation of the German one with four product pages. Your competitor has forty substantial English application pages, and the model notices.

Language gap · same question, two languages · placeholder data
Track A · German
DE

„Welche Hersteller für Blisterverpackungsanlagen empfehlen Sie?“

  1. 01Supplier A
  2. 02Supplier B
  3. 03Your company
  4. 04Supplier C

Named in position 03

Track B · English
EN

“Which manufacturers of blister packaging lines do you recommend?”

  1. 01Supplier A
  2. 02Supplier C
  3. 03Supplier D
  4. 04Supplier E

Your company: not named

Whether the gap between your German answer and your English one is one grade or three, you will know after the check. We measure both separately and show you the difference. For a company with sixty percent export, that is the only number that matters.

Ask any provider how many languages they measure before you sign. The common tools measure one as standard.

Three sentences we hear every week.

“Our customers don’t google. They know us.”

Your thirty existing customers know you, and you are not going to lose them quickly. The three hundred who do not know you yet start from zero — and most of them now start in a chat window.

“We sell through agents and trade fairs.”

Your agent in Malaysia gets called by a buyer who did his research first. The AI answer does not sit behind your sales channel, it sits in front of it — it decides which three numbers get dialled at all.

“We have led our niche since 1923.”

Market leadership is in your accounts. It is often not in the English-language sources these answers are assembled from — and what is there instead, you have never read.

  • 43 %

    Even for a German-language question, the model researches predominantly in English. Your German content does not compete there.

    Peec AI, 10m prompts, February 2026

  • 51 %

    One in two B2B buyers now starts the supplier search in a chat window rather than on Google.

    G2 Buyer Behavior Report, March 2026

  • −59 %

    Google's first result loses this much click-through once an AI answer sits above it. Good rankings no longer protect you.

    Sistrix, 100m German keywords, February 2026

The free check
Five probes

Five questions about your own company you cannot answer today.

  1. “Are we even on the list?”

    We ask, with no prompt about you, for the leading suppliers in your category — the way a buyer who doesn’t know you would ask.

  2. “And on the English list?”

    The same question in English, from the point of view of one of your export markets. That is the list your next customer is reading — and the two lists are rarely the same.

  3. “Is what AI says about us true?”

    First the self-description, quoted verbatim and unedited. Then four to six individual claims from it, checked against your own documentation: correct, incorrect, or unsupported. In our experience this is the page that gets passed around the management floor.

  4. “Who are we losing to?”

    You against your strongest competitor, forced into a direct comparison. The AI has to commit, and you read its reasoning.

  5. “Where does AI send a buyer who doesn’t know us?”

    We describe a typical buyer in your target market with a specific requirement and ask the AI to recommend a supplier. That is the moment an enquiry either happens or doesn’t.

Answer these five for us

Freeten a monthresults in 72 hours

Measure
Enquiries

How you’ll know whether this is working.

Visibility is an intermediate measure. It appears in no set of accounts, and an agency that invents its own metric can also move it.

The measure we let ourselves be judged on is an enquiry from a company that was not in your CRM before.

What isn’t clean about it

A buyer who asked a chat window first and then googled your name will write down Google. We do not claim we can separate those. We claim that a dirty number about enquiries is worth more than a clean number about visibility.

Girdharee Saran
Girdharee Saran · Empfehlwerk
Who does this

Girdharee Saran.

I run every check myself, write every report myself and sit in every conversation myself. No sales team, no pre-qualification, no junior presenting the result.

I have spent twenty years on how companies get found in search systems. The channel has changed twice in that time; the work on the sources has not. Four of those years at Mercer in Vancouver, responsible for English-language content and its localisation for international markets. After that, head of marketing at three companies whose demand is created entirely in English.

That is precisely the layer on which your next export enquiry is decided — and the layer on which most German industrial websites are weakest.

What I am not: a mechanical engineer

I check what the models say against your own documentation and against the sources the answers are built from — not against a gut feeling for your industry. Where a statement requires a technical judgement, it appears in the report as a question for you, not as a finding. That is not modesty, it is the method.

Where I am based

My company is registered in India. The full address is in the Impressum, and I would rather write it here than have you go looking for it there. What it means for your data is in the privacy notice: transfer on the basis of the EU Standard Contractual Clauses, no disclosure to third parties, deletion on request. Authorised recipient for service in Germany: [name, city].

If I am unavailable

Your prompts, raw answers and measurement logs belong to you and can be requested in full at any time — as a folder, not as a dashboard export. No automatic renewal. If I am unavailable for more than four weeks you may terminate without notice, and unperformed months are refunded. This is in the contract, not only on this page.

Why it is free

Because showing you what I measure beats explaining that I can measure. Several agencies now offer a free check. Mine differs in two ways: it measures two languages, and it checks whether the claims are true. Ten a month — not because more is impossible, but because fact-checking does not go faster, and a report without it is worthless.

Checkable facts

Experience
Since 2006 · two languages daily
Roles
Mercer (Vancouver) · Sapient · iZenBridge · NeuroSensum · Whizlabs
Qualifications
MBA (PGPM), MDI Gurugram · CXL Certified Optimizer
Checkable
linkedin.com/in/giri-saran

What we don’t sell.

Guaranteed mentions.
AI answers are not deterministic. Anyone guaranteeing you a mention is selling you chance.
The title “leading agency”.
The field is not two years old. Anyone claiming to be number one in it either hasn’t counted or is hoping you won’t.
Your phone number as payment.
The consent box under our form covers exactly one thing: that I send you the report. No call sequence, no newsletter, no sharing.
llms.txt as an insider tip.
137,000 domains analysed, no measurable effect on citations. We will create the file and not invoice you for it.
Dashboards as a product.
Monitoring is a tool, not an outcome.
The full list is on the pricing page
Comparison
Including us

Five questions to ask any provider. Including us.

Since spring, several German agencies have started offering this. That is more good than bad for you — it just means you have to compare, and the brochures will look alike. These five questions separate them.

  1. How many languages and markets are measured, included in the price?

    One is standard. If you sell into forty countries, a provider measuring one language is measuring the wrong half.

  2. Do they check whether what is said about you is true — or only whether you appear?

    Those are two different jobs. The second is considerably easier.

  3. What is success measured against?

    If the answer is an index of their own construction, check who calculates it.

  4. Who owns the sources they propose to optimise into?

    A trade publisher selling you placement in its own title has two interests. That need not be a bad thing. You should simply know.

  5. What exactly does the consent box under the free check cover?

    If it also covers phone contact, the check was not free. You paid with your number.

Our own answers to all five, at length and including the uncomfortable parts: The guide

Four steps. Start wherever it makes sense for you.

  1. AI visibility check

    You want to know whether this concerns you at all.

    ten a month

    Five probes, three systems, two languages separately. Three pages. 72 hours. Ends with the actions, not with a price.

    free

    Request it

  2. AI visibility audit

    You want a defensible basis for a budget decision.

    50 to 100 questions, three systems, German and English evaluated separately, every claim about you checked against the source it came from. Competitor comparison, source map, action plan with effort estimates. Four to five working days. A 60-minute results conversation, with me.

    Credited

    Book a sprint within 30 days of the results conversation and the €3,500 is credited in full. Book nothing and you keep the report.

    €3,500

    An audit that measures one language costs less. It also measures half.

    Discuss the audit

  3. Visibility sprint

    The gaps are known and should be closed.

    Work on the sources actually cited in your answers: English application pages, trade media, directories, structured data. Monthly measurement in both languages.

    Three months minimum. Monthly cancellation thereafter, no automatic renewal.

    €5,000

    per month

    Discuss the sprint

  4. Ongoing

    What was built should hold when the models change.

    Six months, because source work under six months produces nothing you could honestly evaluate.

    Monthly cancellation thereafter, no automatic twelve-month renewal.

    from €3,000

    per month

Prices

All prices net. Service contract with a written scope.

Terms

Three months for the sprint, six for ongoing work, monthly thereafter. No silent renewal. No setup fee.

No success fee on mentions

With non-deterministic systems that would be a bet I sell you and chance settles.

Timing
Two reasons

Why this doesn’t wait until spring.

Two reasons, both checkable. Neither of them is a discount expiring tomorrow.

Editorial deadlines.

The sources these answers are built from — market overviews in trade media, industry directories, substantial English application pages — run on eight to twelve week lead times. Measure in September and January can look different. Start in January and you are talking about next summer.

The company next door.

86 % of German industrial marketing leaders expect this to be a must-have within two years (bvik Trendbarometer 2026, n = 282). Since spring, several German agencies have added it to their offering. That does not mean you have to hire one of them. It means your competitor is being called in the same quarter you are.

bvik Trendbarometer 2026, n = 282

In 72 hours you will know whether you have a problem.

Five probes, three AI systems, two languages measured separately, every claim held against a source, three pages. If there is nothing there, the report says so — and you have certainty instead of a hunch.

Show me what AI says about us

Free.Ten a month.Every report made by hand.