Dentsu compared on what its own US site publishes, and the footer line that tells you which entity you are hiring

If you are a US marketer shortlisting Dentsu, the first thing to establish is which Dentsu. On 21 October 2026 we fetched dentsu.com and were served a page titled "dentsu APAC". We then fetched dentsu.com/us and were served a page titled "dentsu USA" with a header reading "United States". Same domain, same afternoon, two different regional entities with different brand rosters. Our requests leave from an Indian IP address, so the first fetch is the geography talking rather than the company. The practical consequence is simple and it is the point of this page: read dentsu.com/us, not dentsu.com, and do not accept a credential deck sourced from the global group as a statement about the team that would service your account.

The criterion, applied to every network here including us

The criterion on this page is what the company publishes on its own site that a buyer can verify before briefing it. For a holding group that means four things: a scale figure with a unit, the named operating brands you could actually be assigned to, who the client-facing entity is, and the proprietary platform it will put in the plan. Nothing here is a judgement on creative quality, which cannot be ranked on anything a reader can check, and this page does not try.

What dentsu.com/us publishes, as read on 21 October 2026

The self-description is "Dentsu is an integrated growth and transformation partner to the world's leading organizations." The company was "Founded in 1901 in Tokyo, Japan", and the page carries a heading reading "Over 120 Years of Innovation and Earning Client Trust". That second line is a tenure claim rather than a metric and should not be read as one.

The one scale figure on the page is "over 145 countries and regions". It describes the global group, not the US business. No employee count, client count or revenue figure appears, so the US entity publishes no size of its own that a buyer can check.

The brands it names as "global leadership brands" are Dentsu Creative, Tag, Carat, iProspect, dentsu X and Merkle. Six. The Asia Pacific page we were served first listed Carat, Dentsu B2B, dentsu creative, Dentsu x, iProspect, Merkle and tag, which is seven, with Dentsu B2B present there and absent from the US six. Both renderings are live on the same domain and we are reporting both rather than picking the tidier list.

The service structure is published as "integrating our services across Media, CXM and Creative", with two named approaches, "end-to-end experience transformation (EX)" and "business transformation (BX)".

The proprietary platform is Merkury, which the page describes as "the pioneering (and patented) person-based identity and data platform". That is the sentence in the plan you should ask the hardest questions about, because an identity platform is where a media relationship becomes difficult to leave.

The footer line, which is the most useful thing on the page

The US site carries a legal statement in its footer that opens "Notwithstanding any actions or statements by the Dentsu Group globally or outside the U.S.," and goes on to say that it "will continue to abide by all applicable laws and regulations in the U.S.", with statements relating to the Dentsu Group that are inconsistent with US law to be "interpreted as, and deemed, not applicable to the U.S."

We are quoting what renders and not interpreting it. What it tells a buyer operationally is clear enough without interpretation: the US business is presenting itself as a distinct entity whose commitments are its own, separate from statements made by the group elsewhere. So when you brief this network, establish in writing which legal entity is contracting, which named brand from the list above is delivering, and whether any claim in the pitch originates with the global group or with the entity signing. That is good practice with any multinational network and this one has put the reason on its own homepage.

How to read it against the networks beside it

Every global network is a holding group with operating brands underneath, and the brand you get assigned is a bigger determinant of your experience than the group you picked. The comparison worth making is therefore between the operating brands rather than between the parent names.

We have read several of these the same way. Ogilvy, compared with the networks beside it on what its own site publishes about scale, is the nearest comparison for a creative-led brief, and the Publicis Groupe roster and how that group is organised is the one to read on structure. We measured our own internal linking on 21 October 2026, and 1,015 of our 2,598 live pages have no durable inbound link at all. The Ogilvy page had none when we measured it that morning and had collected two by the afternoon, from pages published while this one was being written, which is the reason we re-measure this kind of claim rather than letting the first reading stand.

Two practical notes from reading these sites side by side. First, brand names in this industry change under you, so verify the name in your pitch deck on the delivering brand's own site before you sign; we have found four renames this month that way. Second, Dentsu's own name does not render consistently, appearing as lowercase "dentsu" in the logo and alt text and capitalised "Dentsu" through the body copy on the same page, which is worth knowing only so that you do not read a style difference as two different companies.

The four questions to put to any network, including this one

• Which named brand is delivering, and which entity is contracting. Get both in the proposal, not the parent name.
• What does the US business employ, and on my account. The published figure covers 145 countries and regions. Yours covers one team.
• What happens to the data in the identity platform if we leave. Ask before Merkury is in the plan, not after.
• Which of the published claims is the group's and which is yours. The footer invites this question, so ask it.

The channel a network of this size is not reporting on

A global network will report media delivery, reach, frequency, search position, brand tracking and attribution. It is very good at all of that. There is one place a buyer now makes a decision that none of it covers.

When somebody asks ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews who they should buy from in your category, the answer is composed for that person at that moment, and it names a few companies. There is no placement to buy, no position to report and no impression to count, so it does not appear in a media report at all. A brand can be running a flawless campaign through any of the brands listed above and simply not be one of the names that comes back.

AI Knows Us measures that and nothing else. We are not an agency, we are not a network and we are not an alternative to one; if you need media, creative or CXM at this scale, hire from the list above. What we do is ask your buyer's question in all five apps on a schedule and record which companies were named, which is also the one number worth adding to a network's reporting line, because it is the only part of the funnel nobody in the pitch is currently accountable for. From our own pricing page read 21 October 2026, one credit is one buyer question on one app once, and "A Claude check uses 3, because Claude costs us more to run." The cheapest plan that reads all five surfaces is Growth at $299 a month with 100 buyer questions watched and 1,200 scheduled answers a month. One credit is one buyer question, asked on one AI app, once, and a Claude check uses 3; the pricing page sets Growth at 1,200 scheduled answers a month. The $149 SEO Start tier carries no AI answer checks.

Where a rival beats us, named, and it matters at this scale. Profound, read at tryprofound.com/pricing on 21 October 2026, publishes "Up to 9 Answer Engines tracked" on Enterprise against the five surfaces we read, "SSO/SAML + SOC 2 compliance", and "Both plans include unlimited seats". We searched our own pricing page's HTML on word boundaries the same day and found no SSO, no SAML, no SOC 2, no ISO, no white label and no DPA in it. A brand buying through a global network will have a procurement and security process that asks for every one of those, and on that axis Profound is further along than we are today. Our own pricing page also renders "More than one business" as No across all eight plans while a separate row lists 10 businesses covered for Agency, which for a multi-brand advertiser is a contradiction worth resolving with us before buying. Both lines render.

Questions about briefing Dentsu in the United States

Which Dentsu brand would actually run my account

One of the six the US site names: Dentsu Creative, Tag, Carat, iProspect, dentsu X or Merkle. Media briefs tend toward Carat and iProspect, creative toward Dentsu Creative, data and CXM toward Merkle. Ask for it in writing, because the group name tells you nothing about who shows up.

Why did dentsu.com show me a different region

Because the domain routes on where the request comes from. Ours came from an Indian IP and we got the Asia Pacific page. Go to dentsu.com/us directly, confirm the header reads United States, and take your figures from there.

How big is Dentsu in the US

Its US page does not say. The only published scale figure is "over 145 countries and regions" for the global group. If US headcount matters to your brief, it is a question for the pitch rather than something you can verify first.

What is Merkury and should it worry me

The site describes it as "the pioneering (and patented) person-based identity and data platform". Treat it as you would any identity layer: useful, and a thing that makes leaving harder. Settle data ownership and exit terms before it is in the plan.

Is a holding group a better buy than an independent agency

For multi-market media at volume, the group's buying scale is a real advantage. For a single-market brief where senior attention is what you are short of, an independent often gives you more of the people you met in the pitch. Decide on which of those two you actually need.