Jellyfish, the agency: what it sells, the named platforms behind the pitch, and how to read the figures on its own site

Jellyfish describes itself as the integrated global digital marketing business, uniting media, creative and data through technology and generative artificial intelligence. The useful thing to understand before a first meeting is that the offer has two halves: an agency service across media, creative, data and strategy, and a set of named platforms the agency has built and trademarked. A buyer is choosing both at once, which is an advantage when the platforms do something you need and a complication when you want the service without the stack. The list below is read from the company's own site, and its published performance figures are presented as the company's claims rather than as independent findings.

A disclosure that belongs at the top rather than the bottom: one of those platforms, Share of Model, measures how brands are perceived by large language models, which is the same lane the company publishing this page works in. Read this as a competitor's account of a competitor, check it against jellyfish.com, and weigh it accordingly.

What it sells, as listed on its own site

The expertise areas published on the site are AI solutions, creative, data, media, partner work, technology, strategy and training. That spread is the definition of an integrated agency: media buying and creative production under one roof, with a data and technology practice attached and a training arm beside it.

Two consequences for a buyer. An integrated shop can be briefed once for a campaign that runs across channels, which removes the coordination problem that comes with three specialist suppliers. And the same structure makes it harder to tell what you are paying for per discipline, which is why the fee model matters more here than with a single-service supplier. How agency fees get built, and the clauses that decide ownership, is set out in how an advertising agency fee is built.

The named platforms

Platform	What the site says it does
Share of Model	Tracks how a brand is perceived across large language models, to show visibility, positioning and competitive standing in those answers
Social Agents	Tracks cultural signals in real time, to catch trends before they are obvious
Switchboard	Connects creative and media performance data, so the two can be read together
Now Next Soon	Runs simulations of a media plan to map outcomes before budget is committed
J plus Bidding	Adjusts in-market budget so each spend informs the next
J plus Scribe	Refines generated text for product descriptions and messaging
AI Studios	Production at volume, with asset creation handled by the platform
Pencil	The creative platform behind the production claims, described on the site as trained on 2.65 billion dollars of ad spend and 235,000 generated assets
Creative Collective	The group the site credits with original ideas rather than ideas drawn from common training sets
Language Services	Creation and adaptation of content in more than 150 languages

Trademarked platform names are worth noting as names rather than as categories. Share of Model is Jellyfish's product in the AI-visibility category, not the industry's term for the measurement, in the same way that one vendor's authority score is not the industry's measure of authority.

The figures on the site, and the questions to put to them

The site publishes a set of performance numbers: assets created 55 per cent faster, content delivered 80 per cent faster, revenue up 19 per cent on 14 per cent less budget, sales up 79 per cent, greenhouse gases down 27 per cent, and a case headline describing one search strategy driving 19 per cent growth across 38 markets with 14 per cent less budget.

Those are the agency's own published claims, read on 20 October 2026. Four questions make any of them usable in a decision, and they are the questions to ask of every agency including the one writing this.

• Which client, which market, which period? A percentage with no client and no window is a headline rather than a result.
• Against what baseline? Faster than their previous process, or faster than an industry norm, are different claims with the same number.
• What else changed at the same time? A 19 per cent rise during a product launch or a category boom is not the same evidence as one during a flat quarter.
• Can we speak to that client? The answer to this question sorts proposals faster than the numbers in them.

Who it fits

Fits well	Fits badly
Multi-market advertisers with real media budget to place	A local business needing to be found in one city
Brands wanting media, creative and data briefed once	A buyer who wants one specialist discipline and nothing else
Teams who will actually use a platform layer	Teams with their own stack they intend to keep
Advertisers running in many languages	Single-market advertisers with one language
Organisations that can supply the data the tooling needs	Organisations whose data lives in places nobody can export from

What to settle before signing with any global agency

• Who is on the account, by name, and what share of their week you get. On a global roster this is the question with the most variable answer.
• The fee model. Retainer, project, commission on media, or a platform licence inside the fee. Ask which parts are which, in writing.
• Who owns the data and the accounts. Ad accounts, analytics properties, creative files and any audience lists should be yours, not held inside the agency's structures.
• What happens to the platform access at the end. Proprietary tooling is a reason to buy and a switching cost to understand, and the exit terms are the place that gets settled.
• Where the money goes. On media buying, ask what share of your spend reaches a publisher, which is the transparency question covered in buying transparency terms from a media agency.
• How it compares with a holding company roster, where the equivalent capabilities sit in separate named agencies under one group, as set out in how a holding company roster is organised.

Questions buyers ask about Jellyfish

What does Jellyfish do?

It is an integrated global digital marketing business: media planning and buying, creative, data, technology and strategy, with a training arm and a set of its own named platforms. Its own description is the integrated global digital marketing business uniting media, creative and data through technology and generative AI.

Is it an agency or a software company?

An agency with productised tooling. The platforms are built and named by the agency and come as part of the engagement rather than as separate shelf software, which is worth confirming in the fee conversation.

What is Share of Model?

Jellyfish's platform for tracking how a brand appears in large language model answers, described on its site as showing visibility, positioning and competitive standing in those systems. It is one vendor's product in that category rather than a standard metric.

Which markets does it work in?

The site offers United States, United Kingdom, Spain, France and Brazil language editions and advertises content services in more than 150 languages, which is the profile of a multi-market operation.

Is it suitable for a small business?

Generally not. The structure and the tooling are built for advertisers placing meaningful media across markets. A single-location business is better served by a local supplier or a specialist.

How do I compare it with other agencies?

Brief two or three the same way, in named outputs with dates, and compare on who does the work, what the fee covers, who owns the data and what the exit looks like. Platform names should be read as capabilities to test rather than as reasons to decide.

Measuring the thing both of us are selling

The most interesting line on Jellyfish's site is the one that says large language models are now one of your most important audiences, because it is a claim an agency and a measurement company can agree on while disagreeing about everything else. A buyer who accepts it has a practical next step that costs nothing and needs no supplier: ask the question your customers ask, in their words, with your own name left out so the answer arrives clean, and write down which companies get named and which sources the answer leaned on. Do that for four weeks and you own a baseline before anybody sells you a platform. AI Knows Us runs exactly that, across the main assistants, and reports who gets named and what each answer was built from.