An advisory perspective from Happy Cog · Vol. 02 · Updated September 2026
Digital Outlook 2027.
The 2026 problem
Being legible to machines.
The 2027 one
Being transactable by them.
What changed across design, development, and digital marketing in 2026, and what we would do about it in 2027. It opens with a scorecard on last year's forecasts, including our own, because a document that makes predictions should be willing to show how the last round held up.
The outlook in four sentences
Through most of 2026, artificial intelligence read the web, summarized it, and recommended what to buy. By the end of the year it had started doing the buying. Being legible to machines was the 2026 problem, and being transactable by them is the 2027 one. The practical consequence is that the boundaries between building a website, writing content, and running media have dissolved, because a brand's AI visibility now depends on a fast semantic build, content structured to be cited, clean first-party data, and a machine-readable catalog, all at once.
01 · The scorecard
How did last year's forecasts actually hold up?
The industry entered 2026 with a set of confident numbers, and the direction of the misses is more instructive than any single figure. Forecasters were too conservative about spend and too aggressive about how fast automation would deliver. The IAB projected US ad spend growth of 9.5% in January and raised it to 12.3% in September. Google's AI Max claim went from roughly 14% more conversions at beta to a published 7% for the full feature suite, with independent studies disagreeing sharply. The finding that top-10 ranking explains roughly 75% of AI Overview citations has since been read closer to 17% to 38%.
We would treat platform-published performance claims the way we treat a forecast, which is as a directional input with a named source and a margin rather than a number to plan against.
Interactive · Make your call
Each card shows only what was predicted. Guess before you reveal.
Nine forecasts: five from January's industry consensus, four from our own Vol. 01, graded by us.
- Too conservative
- Revised down
- Unresolved
- Held
- Missed
What the January consensus said
IAB · January
US ad spend would grow roughly 9.5%.
What happened by September
The IAB raised its full-year forecast to 12.3% on September 10, after a stronger than expected first half.
IAB · January
Social would grow 14.6% and CTV 13.8%.
What happened by September
Revised up to 16.5% and 15.6% respectively.
eMarketer · April
Meta would overtake Google in net ad revenue.
What happened by September
Still a projection rather than a banked result, on a worldwide gap of roughly $4B.
Google · At beta
AI Max would deliver about 14% more conversions.
What happened by September
Google now publishes an average of 7% for the full feature suite, and independent studies disagree sharply.
Industry analysis
Ranking in the top 10 drives roughly 75% of AI Overview citations.
What happened by September
Later readings cluster far lower, in the range of roughly 17% to 38%.
Graded by usWhat we said in Vol. 01
Happy Cog · Vol. 01 · Design
Machine experience is the idea to watch: a website has a second audience which is not human.
What happened in 2026
In 2026 it stopped being an idea and became a line item.
Happy Cog · Vol. 01 · Paid media
Platform recommendations are designed to increase spend rather than results. Treat them as inputs, not instructions.
What happened in 2026
Ten weeks later Google removed the choice for a large share of Search campaigns, auto-upgrading them to AI Max on September 1.
Happy Cog · Vol. 01 · Search
How much AI Overview citation is explained by ranking is contested, with one analysis near 75% and a re-reading closer to 38%.
What happened in 2026
The lower estimates have more support and some readings go lower still. The relationship is real, and how large it is remains contested.
Happy Cog · Vol. 01 · The thread
AI is becoming the layer between a brand and its audience.
What happened in 2026
That was right, and it is now incomplete. Through 2026 the layer stopped only interpreting and started acting.
Sources: IAB US ad spend outlooks (January and September 10, 2026), eMarketer (April 2026), Google AI Max documentation. Vol. 01 positions are drawn from Digital Outlook 2026.
The full scorecard, and the rest of the evidence, runs 35 pages.
Discipline chapters, portfolio data from 56 managed accounts, and the methodology behind every figure.
Grading ourselves
What we got wrong, and where we still do not know
Two sections of this outlook are flagged as unsettled, because the sources contradict each other and the honest answer is that the magnitude is contested. One of them, agentic commerce, is the least reliable topic in the document, and we say so on the page rather than in a footnote. We also left out several widely repeated figures we could not trace to a primary source.
The sharpest example is our own. Connected TV took $171,573 of managed programmatic spend in 2026 to date, delivering 9.1 million impressions at a 97.7% completion rate. None of those flights ran with a holdout or a geo control, which means we cannot tell you what the spend was worth, and the conversions attribution assigns to it are exactly the over-crediting we warn about elsewhere. We are reporting the gap rather than working around it, and closing it is the first thing on our own 2027 list.
From our portfolio · Yahoo DSP connected TV · Jan 1 – Sep 22, 2026
$171,573
CTV spend, 2026 to date
97.7%
Completion rate on 9.1M impressions
0
Flights that carried a holdout
Delivery at an $18.94 CPM, roughly three quarters of all DSP video investment. Attributed conversion figures are deliberately omitted. From 2027, every CTV flight we run ships with a holdout or geo test.
02 · Executive summary
What actually changed in 2026
-
01
The market ran hotter than the forecasts.
Social grew 16.5%, connected TV 15.6%, and paid search 8.1%, roughly half the rate of social (IAB, September revision).
-
02
Automation stopped being a choice.
Google auto-upgraded Search campaigns to AI Max on September 1 with defaults most advertisers never reviewed. Dynamic Search Ads follow in February 2027.
-
03
Advertising moved inside the assistants at commercial scale.
ChatGPT Ads reached a reported $1B annualized run rate in under 200 days (OpenAI), with self-serve buying open across dozens of markets.
-
04
Product data became a distribution channel.
A catalog that is not machine-readable is invisible to every competing commerce protocol.
-
05
Measurement became the bottleneck.
The IAB found 86% of buyers changing how they measure because of conversational AI and 45% unable to reconcile AI-driven and traditional journeys.
03 · The macro picture
Where should the 2027 budget actually go?
The single most consequential fact about 2027 is that the cyclical lift disappears. The Olympics, the World Cup, and the midterms all sat inside 2026, so anyone building a 2027 plan against 2026 comparables is measuring against an inflated base. Forecasts for 2027 global growth run somewhere between 5% and 8% depending on the house, and we would show a client that spread rather than a single number, because the range reflects real disagreement.
In our own book the budget is following the automation. Performance Max rose from 38.7% to 45.3% of measured spend while Search fell from 52.7% to 44.9%, making Performance Max the largest single channel for the first time, and standard display spend fell roughly 74% (Happy Cog, 56 managed Google Ads accounts).
5% to 8%
2027 global growth, depending on the house
If your 2027 targets assume you repeat a World Cup year, rebuild them. Strip the event-driven demand out of your 2026 baseline before you set a growth number.
04 – 07 · By discipline
What we would tell each team.
04 · Design
Design
Last year we described machine experience as the idea to watch. In 2026 it became a line item: strict heading hierarchy, semantic markup, and clean component logic are now the difference between a brand an AI can cite and one it cannot see. In 2027, a design will be judged on whether a machine can read it as much as on how it looks.
Happy Cog's take
Brand expression and machine readability are two requirements of the same build, and we design for the person and the crawler in the same pass. In 2027 the crawler is the reader most teams will get wrong, because structure is now measurable and taste is not.
05 · Development
Development
If 2025 reorganized development around AI assistance, 2026 was the year the industry started counting the cost of doing it without governance. Gartner expects more than 40% of agentic AI projects to be cancelled by 2027, citing unclear value, cost overruns, and weak risk management.
Happy Cog's take
Speed without governance is a liability. We let AI compress timelines only behind human review, automated checks, and rollback discipline. When something goes wrong in an AI-assisted build, the question a client will ask is who reviewed it.
06 · Search and organic visibility
Search
Roughly two thirds of US Google searches now end without a click. The IAB found 76% of buyers naming optimization for AI answers their top priority, and 48% now measuring citations inside AI tools.
Happy Cog's take
The work is still SEO, with more asked of it. Clean technical foundations, clear entity structure, and authoritative content are what make a brand citable by a model. On top of that we add structured answers, schema, and passage-level content, which is why “are we part of the AI answer?” is a reporting line for us.
07 · Paid media
Paid media
Google made AI Max mandatory in stages, Meta kept automating end to end, and Google display shrank as Performance Max absorbed its inventory. The money leaving the Google Display Network should go where the controls are, which for us is programmatic.
Happy Cog's take
Paid media is still a precision tool, and automation has not changed that. When a platform removes a control, the work does not disappear. It moves upstream, into the conversion data and the constraints you set before the system starts spending.
09 · The thread that ties it together
The same shift shows up in every discipline
Design now has a second audience in the crawler, and structure is how you speak to it. Development hands much of the build to AI, which makes governance the thing that keeps the product trustworthy. In organic search, AI decides who gets cited, and authority and structure are how you earn the citation. In paid media, AI runs the campaign and steers by whatever data and creative you feed it. In commerce, AI recommends what it can read in your product feed.
| Discipline | What the AI does | How you steer it |
|---|---|---|
| Design | The crawler is a second audience | Structure |
| Development | AI builds much of the product | Governance |
| Organic search | AI decides who gets cited | Authority plus structure |
| Paid media | AI runs the campaign | Your data and creative |
| Commerce | AI recommends the product | Your feed |
As systems take over execution, the brief becomes the product.
Constraint setting, data quality, and creative judgment used to be the preamble to the work, and now they are most of it.
11 · Scored in Vol. 03
Ten calls for 2027, which we will grade next year
Each one is specific enough to be graded a year from now, and each carries a confidence level.
-
01 High confidence
Google completes the Dynamic Search Ads migration to AI Max roughly on the February 2027 schedule, without a further deferral.
-
02 Moderate confidence
At least one of the major agentic commerce protocols consolidates into another or is effectively abandoned.
-
03 High confidence
AI-referred sessions remain below 5% of total sessions for most mid-market brands through 2027, while converting at a materially higher rate than average organic traffic.
-
04 Moderate confidence
AI visibility reporting becomes a standard line in agency reporting rather than a paid add-on.
-
05 High confidence
A sizable share of marketers miss 2027 targets set against inflated 2026 comparables, and the misses cluster in categories that leaned hardest on event-driven demand.
-
06 Lower confidence
A significant public incident involving an under-governed AI agent changes enterprise procurement language faster than analyst guidance has.
-
07 Moderate confidence
Assistant advertising becomes a standard line item in mid-market media plans rather than an experimental budget, with measurement still the main complaint.
-
08 High confidence
Standard display continues its decline and largely disappears from performance-led plans, surviving mainly in programmatic reach buys.
-
09 Moderate confidence
At least one major platform walks back an automation default after advertiser pushback, in the way Google restored Dynamic Search Ads creation in 2026.
-
10 Moderate confidence
Conversion data quality becomes an explicit agency selection criterion, because clients learn the hard way that automation amplifies whatever the tracking tells it.
Want to see how these hold up?
Tick "Tell me when these are graded" on the form at the foot of the page and we will send you the Vol. 03 scorecard when it publishes.
10 · What to prioritize · Interactive
Where we would start
None of these require chasing the newest feature. Each one strengthens fundamentals that every AI surface now rewards. Your progress is saved in this browser only.
Questions
Frequently asked.
How accurate were the 2026 digital advertising forecasts?
Forecasters were too conservative about spend and too aggressive about how fast automation would deliver. The IAB projected US ad spend growth of 9.5% in January and raised it to 12.3% in September. Google's AI Max claim went from roughly 14% more conversions at beta to a published 7% for the full feature suite.
How much did US ad spend grow in 2026?
The IAB put US ad spend growth at 12.3% for 2026 in its September revision, nearly three points above its January projection of 9.5%. Social grew 16.5%, connected TV 15.6%, and commerce media 13.6%, while paid search grew 8.1%, roughly half the rate of social.
How much will ad spend grow in 2027?
Forecasts for 2027 global growth run between roughly 5% and 8% depending on the house. The Olympics, the World Cup, and the US midterms all sat inside 2026, so a 2027 plan built against 2026 comparables is measuring against an inflated base. Strip the event-driven demand out of the 2026 baseline before setting a growth target.
What is the Google AI Max migration?
Google stopped allowing creation of campaign-level broad match and legacy automatically created assets in early August 2026, then auto-upgraded Search campaigns using those settings to AI Max from September 1 with defaults most advertisers never reviewed. The Dynamic Search Ads migration follows in February 2027. Audit conversion actions before campaign settings, because automation optimizes toward whatever the tracking says matters.
Is Performance Max overtaking Search?
In Happy Cog's managed book of 56 Google Ads accounts, Performance Max rose from 38.7% to 45.3% of measured spend between the second half of 2025 and the July to September 2026 window, while Search fell from 52.7% to 44.9%. Performance Max became the largest single channel in that book for the first time, inside nine months.
How should connected TV be measured?
With a holdout or geo test, set up before launch. Nobody clicks a TV, so measure what CTV changes: exposed-household site visits, branded search volume, direct traffic, and conversion rates in other channels. Attribution alone tends to over-credit CTV when search and social are also running.
What is agentic commerce?
Agentic commerce is shopping in which an AI assistant reads structured product data to decide what to recommend and increasingly moves the shopper toward a transaction. Several competing protocols define how agents read catalogs and execute payments, and none has won. A catalog that is not machine-readable is invisible to all of them, so the practical work is fixing the product feed.
What should marketers prioritize in 2027?
Rebuild the 2027 baseline before setting targets, audit conversion data before auditing campaigns, and instrument AI visibility so a trend exists when it matters. Then make the site machine-legible, fix product and service data, and put guardrails around ad automation. None of it requires chasing the newest feature.
Start a conversation
If this outlook describes your year, let's talk about 2027.
Happy Cog has been doing design, development, and digital marketing as one team since 1999. Tell us what you're planning, what's stuck, or which of these calls you think we got wrong, and someone from the team will get back to you.
The full outlook runs 35 pages, with the discipline chapters, the portfolio data, and the methodology behind every figure. No form required.
Download the full outlook