CMOs give AI 15.3%. Only 30% are ready.

Marketing has stopped asking whether to use AI and started asking how much to spend on it. Gartner's 2026 CMO Spend Survey found that chief marketing officers allocate 15.3% of their marketing budgets to AI initiatives, while total marketing budgets barely moved, to 7.8% of company revenue from 7.7%. Only 30% of the 401 marketing leaders surveyed report mature AI readiness. Money is moving faster than capability.

At the same time the customer has moved. Bain estimates that 30% to 45% of US consumers already use generative AI for product research and comparison. A growing share of discovery now happens inside an assistant that summarises, ranks and recommends before a shopper ever reaches a brand's site.

ETHORITY reviewed 24 marketing AI offerings for the Good AI Prism in October 2026, reading each provider's own pricing, security and legal pages. This piece reports what they publish, and what a buyer should ask before signing.

Five jobs, from brief to visibility

The Marketing & Content area map sorts the offerings by job. Content creation and brand voice: Jasper, Writer, Copy.ai, neuroflash, Typeface and Markup AI. Design and assets: Canva Magic Studio, Adobe GenStudio and Bynder AI Agents. Campaign orchestration and messaging: BrazeAI, SAP Engagement Cloud, Optimizely Agent Platform, Jacquard and Persado. Paid media automation: Meta Advantage+, Google Performance Max and Albert. Social and listening: Hootsuite Wisdom, Iris AI and Composer.

A fifth job is new in this edition: visibility in AI answers. Peec AI, the AI Visibility Toolkit, Surfer and BrightEdge Copilot measure or improve how a brand appears in assistant answers, not only in search results. It is the market's response to the Bain numbers, and it is still young enough that method matters more than dashboards. Ask each provider which assistants it samples, how often, and with which prompts.

Prices you can see, units you must compare

Ten of the 24 publish a price. The units range from seats to tokens. Jasper lists its Pro plan at US$69 per seat per month, or US$59 billed yearly. Canva Pro costs US$12 per person per month billed yearly. neuroflash starts at €42 a month billed annually. Writer publishes API prices per million tokens, for example US$2 input and US$8 output for Palmyra X6. Hootsuite lists from US$99 per user per month.

Seat prices are easy to budget and easy to waste, because licences outlive enthusiasm. Token and credit prices follow usage and can spike in a campaign month. Paid media tools such as Advantage+ and Performance Max carry no licence fee at all; the cost sits in the media spend and in the control you hand to the platform. Compare on cost per approved asset or per conversion, not on the list price.

15.3% of marketing budgets now go to AI initiatives, according to Gartner's 2026 CMO Spend Survey. Only 30% of CMOs report mature AI readiness.

Scenario. Five Jasper Pro seats cost US$345 a month, or US$295 at the yearly rate of US$59. Canva Pro for the same five people costs US$60 a month billed yearly. Neither figure means much until it is divided by approved assets or conversions.

Brand data, training and where it lives

Marketing tools hold the assets that make a brand: voice guides, images, customer segments and campaign results. Training terms decide who else learns from them. Twelve of 24 document that customer data is not used for training, or only with opt-in. Some state that it can be used. Hootsuite's AI terms allow use of customer data for training. Typeface fine-tunes on a customer's content and brand assets for that customer's dedicated models, which is often exactly what a brand wants, provided the contract says the models stay dedicated.

Hosting is thinner still. Seven of 24 document an EU hosting or data-residency option, among them Bynder with an EU-only option in Frankfurt and neuroflash with servers in Germany. Braze states that EU data routing is not implemented for its Operator assistant. For a European brand with customer segments in the prompt, that is a contract question, not a footnote.

30–45% of US consumers already use generative AI for product research and comparison (Bain & Company, Nov 2025).

The labelling deadline

AI-generated content now has dates attached. Under the AI Act, as amended by Regulation (EU) 2026/1744, the Article 50 transparency duties apply since 2 Aug 2026. Providers of generative systems must mark synthetic outputs in a machine-readable format, and systems already on the market before that date have until 2 Dec 2026. Deployers that publish deepfakes must disclose them. Our article on synthetic faces and voices covers the rules in the EU, China, India and South Korea.

For marketing teams this means two checks. Does each image, video and voice tool mark its output? And does the publishing workflow keep the marks, or strip them in export and compression?

How to run a 30-day test

A fair test of a marketing AI tool takes a month, one channel and two teams. Pick a channel with steady volume, such as product pages, newsletters or paid social. Split the work: one team produces with the tool, one without, both under the same brief and the same approval rules. Record four numbers for each asset: time from brief to approval, rounds of revision, cost including licences and agency hours, and the channel result, whether click-through, conversion or reach.

Two rules for an honest test

Two rules keep the test honest. First, count rejected outputs as cost, because a tool that drafts fast but needs heavy rework is slower than it looks. Second, keep the legal check in the loop: who owns the output, which licence covers images and fonts, and whether the tool marks synthetic content. A test that ignores rights saves time now and spends it later in takedowns.

At the end, the decision is simple. Scale the tool where the numbers moved, drop it where they did not, and write down why. That record becomes the evidence base for the next budget round, which is exactly what the readiness gap in the Gartner survey says most teams lack.

What the readiness gap means

In early 2025, 27% of CMOs reported limited or no GenAI adoption in their teams. A year later budgets have moved, but readiness has not caught up. The Global Market Study's advice for this function is plain: choose tools by job, check IP and licence terms, and measure content cost per asset and conversion. That beats buying a platform for every job and hoping usage follows.

What to do Monday

1. Inventory every AI tool already in use. Include free and personal accounts; brand assets often sit there.
2. Pick one job and one measure. Cost per approved asset, time to publish or conversion rate for one channel.
3. Shortlist three per job. Use the area map, and keep visibility tools separate from content tools.
4. Ask four questions in writing. Price unit, training on your assets, hosting region and output marking.
5. Set the labelling rule. Decide by 2 Dec 2026 how synthetic images, video and voice are marked and disclosed.

Retail teams should read our piece on AI in pricing and personalisation. For the hand-over from campaign to pipeline, see the Sales & CRM topic area.