When the model names a forwarder, are you in the sentence?
Shippers now ask ChatGPT, Claude and Perplexity who to use. Most logistics and construction brands never appear. We measure that, then do the work that puts the operator in the answer.
The shortlist is now a generated paragraph.
A procurement lead asks a model for a facade logistics partner from Dubai to Melbourne, or a forwarder who can handle project cargo into a live site. The answer names two global brands and a company that has never done the job. The specialist, the one who actually runs that trade, is not in the paragraph. They have a website. They have twenty years. The model has never been given a reason to say their name.
If you are not cited, you are not on the list.
This is not a vanity ranking. It is whether the buyer ever hears of you. Paid search does not cover a question typed into Claude. A directory listing does not survive a model that wants a sentence it can stand behind. Operators who only exist in a brochure and a relationship are invisible the moment the relationship is not in the room.
A programme that measures how generative systems talk about a brand, then changes what they can say.
We work with logistics and construction companies on AI visibility: whether ChatGPT, Claude, Perplexity and Gemini mention them, what they say, and whether that sentence is usable. Then we do the unglamorous work that changes the result: entity consistency, pages a model can cite, operator-authored proof, the technical record of what the company actually does. Not a content mill. The same discipline we bring to an operational system: the object has to be true, or it will not hold.
How it works.
Ask the questions the buyer actually asks.
Not brand terms. The job: facade logistics Dubai–Melbourne, project cargo into a live site, a 3PL who can recover waiting time. We record what each major model answers, who it names, and what it claims about them.
Score cited, recommended, or absent.
Mentioned in passing is not the same as recommended for the job. Absent is the common case. We keep the trail so a quarter later we are not arguing from memory.
Build the record a model can stand on.
Clear entities, pages that state the trade and the geography, proof written by the operator (jobs, lanes, constraints), not a blog that says 'we care about logistics.' Models retrieve and summarise. They need a source that does not collapse when quoted.
Measure again on the same questions.
The programme is a loop. If the sentence has not changed, the work is not done. We do not sell a one-off audit and leave.
What is in the system.
The question set
Built from how shippers and contractors actually enquire: trade, geography, job type, not from the brand's own slogans.
Citation ledger
Per model, per question: named, recommended, misattributed, or missing. The thing a director can read without a briefing.
The citable record
Pages and statements a model can quote without inventing a capability the company does not have. If it is not true on the floor, it does not go up.
Re-measurement
The same questions, later. Visibility that cannot be shown twice is a campaign, not a position.
What this replaces.
Hoping the website is enough
A brochure site is not a source. Models skip it, or they invent from a competitor who wrote the sentence first.
SEO as a proxy
Ten blue links are not a generated recommendation. Ranking for a keyword does not put you in Claude's answer.
A one-off 'AI audit'
A slide of screenshots with no second pass. If you cannot re-ask the question, you did not build a position.
We do not write filler until a model repeats it.
Generative systems are already full of logistics copy that could apply to anyone. That copy is why specialists are invisible: the model has a generic paragraph and no reason to name a real operator. The work is to make the company specific: the trade, the geography, the job, so a citation is a fact, not a hallucination we got lucky with.
ChatGPT, Claude, Perplexity, Gemini
The four a shipper or contractor will actually open. We do not chase an obscure engine to pad a chart.
Specific or it does not count
'Logistics solutions worldwide' is how you stay absent. Facade packages, a corridor, a site constraint: that is a sentence a model can keep.
The floor still has to be true
We will not get a company cited for a trade they do not run. The programme sits on the same honesty as the operational work.
If the model cannot name you, the buyer never will.
We run this with logistics and construction companies who already do the work and are tired of watching a generalist get the sentence. Bring the questions your customers are already asking a model. We will tell you whether you are in the answer.
Book a scoping call