Win the answer, before the answer has a price.

It's real, and it's early. OpenAI began testing ads in ChatGPT on 9 February 2026 — US-only, for logged-in adults on the Free and Go tiers, with Plus, Pro, Business, Enterprise and Education staying ad-free. Placements are clearly labelled 'Sponsored', kept separate from the answer, and OpenAI states they don't influence the answer itself; advertisers get only aggregate views and clicks, not user data. It's worth being precise: this is a weeks-old test that OpenAI nearly shelved during its December 2025 'Code Red', so we treat it as something to test surgically, not as guaranteed inventory. Meanwhile Microsoft Copilot is broadly live and Google's AI Mode formats are mostly 2026 pilots — we run what's live today and stand ready as the rest opens.

It's emerging, and we'll never pretend otherwise — that's exactly why how you enter matters. We don't move your core budget; we run a contained, capped test alongside your proven channels, with a clear kill switch and a profit-led readout. The real return at this stage is learning that can't be bought later: the CPMs, the queries that trigger you, the creative that converts inside a conversation, and the measurement plumbing. Brands that wait don't avoid the cost — they inherit a crowded auction and a playbook a year behind the brands who started now. And much of the highest-leverage work — feed readiness, citations, measurement — pays off on every AI surface regardless of how the ad formats evolve.

Get ready for agentic commerce — it's the safest, most concrete investment on this whole channel. As AI agents start to research, recommend and buy on a user's behalf, the brands that get chosen are the ones with clean, structured, machine-readable product feeds and presence on the emerging open standards: the Agentic Commerce Protocol (ACP, OpenAI and Stripe) and the Universal Commerce Protocol (UCP, Google and Microsoft). This is live now — in-chat checkout exists in ChatGPT and Copilot and is rolling into Google's AI Mode — and feed quality is becoming the new shelf-eligibility: brands with poor feeds simply won't be surfaced or bought. We get your feed and product entities AI-shop-ready before the agent does the buying for your customer.

No — and it's important to be clear about that. As of early 2026 no AI platform sells a paid placement inside the generative answer itself. Ads sit beside or beneath the answer and are clearly labelled; the brand the assistant recommends and cites organically is earned through genuine authority — structured content, schema, entity strength and cross-platform presence, which is the AIO/GEO work we specialise in. We use paid for fast reach and channel signal, and feed what we learn into the earned position that actually compounds. Anyone implying ad spend buys a recommendation inside the answer is misleading you.

With the same profit-led scoreboard we run everywhere, adapted to the surface. The first problem is 'dark AI' traffic: ChatGPT and other assistants often don't pass a clean referral tag, so without proper setup most AI-driven visits get mislogged as 'direct' — studies suggest brands under-count AI traffic by 30–40% on default settings. We fix that by wiring a dedicated GA4 AI-referral channel for ChatGPT, Perplexity, Claude, Gemini and Copilot, instrument conversational-ad clicks, and track share-of-recommendation — how often the assistant names you in your category. For the ad surfaces themselves, OpenAI gives only aggregate views and clicks, so we're honest about the limits and judge the channel on early learning and incrementality, not last-click ROAS.

It means the market is still finding its shape, which is precisely the case for a partner reading it daily rather than a fixed annual plan. Perplexity abandoned ads entirely in February 2026 to protect answer trust; Google keeps ads out of the Gemini chat experience itself; Anthropic refuses ads in Claude — while OpenAI ramps its ad test and Google runs ads in AI Mode. We treat each surface on its merits: lean into what's live and converting, stay out of what's pulling back, and never lock you into inventory that could disappear. Reading that volatility for you, and moving fast when a surface shifts, is the service.

This suits high-spend brands already investing meaningfully across Google, Paid Social or Amazon — typically £10k–£500k+ a month in total media — who can fund a disciplined experiment for first-mover advantage. You don't need a large incremental budget to start: a contained test plus the AIO/GEO and feed-readiness groundwork is often the smartest first step, because that earned and structured layer pays off across every AI surface regardless of how the paid formats evolve. We start from a free, no-obligation readiness audit. And we'll tell you honestly when the channel isn't yet worth scaling for your category, and when it is.

Because we were built AI-first, not retrofitted — this is the rare channel where that's literally true to the surface, not a slogan. Most agencies were built for ten blue links and are bolting 'AI advertising' onto a legacy search retainer, learning the channel on your budget. The same DNA that built our AIO/GEO citation engine makes an AI-native paid surface home turf, and we connect paid placements, earned recommendation and agentic-commerce readiness into one engine an add-on can't replicate. You're hiring the team already building for this channel — running the live formats with discipline and honest about what's real and what isn't — rather than the team catching up to it.