A practical checklist of questions to ask before you choose a marketing agency, covering AEO, incrementality testing, creative velocity, and AI workflows.
A CMO we know at a 40-person B2B SaaS company ran a pitch process last autumn. Three agencies, same brief, same budget range. Two of them walked in with decks built around last year's playbook: more content, more channels, more dashboards. The third opened with a question about how the company's pipeline attribution broke down between first-touch and self-reported sources. That agency won the account before the second meeting.
The lesson is not that the third agency had better slides. It is that they had a working model of where marketing is actually heading, and they tested it against the client's numbers instead of reciting it. When you choose a marketing agency in 2026, that is the thing you are really buying: a point of view you can stress-test, not a service menu.
Here is how to run that stress test.
## Why the 2026 agency shortlist looks different
Search behavior is fragmenting. Buyers increasingly get answers from AI assistants, not blue links, which pushes agencies toward answer engine optimization (AEO) alongside classic SEO. Paid social costs keep climbing while creative volume requirements double. Attribution is moving from last-click dashboards to incrementality testing and marketing mix modeling, because cookie-based tracking keeps degrading.
None of this is speculative. It is already showing up in pitch decks, media plans, and reporting templates. The question is whether the agency across the table has adapted its operating model or just its vocabulary.
## The questions worth asking
Treat this as a checklist. Ask all of them, then watch how the answers change under follow-up.
### 1. "How do you handle answer engine optimization for a client like us?"
A weak answer: "We optimize for featured snippets."
A strong answer names the mechanics. They will talk about structuring content so AI assistants can cite it, monitoring brand mentions inside AI-generated answers, and tracking zero-click visibility separately from organic sessions. Ask them to show a client where AI referral traffic or branded AI citations were measured. If they cannot produce a number, they are guessing.
### 2. "What percentage of your paid media budget went to incrementality testing last year?"
This separates agencies that still trust platform-reported ROAS from ones that verify it. Platform ROAS is a starting point, not proof. An agency worth hiring will describe holdout tests, geo experiments, or at minimum a plan to run them. Ask what they did when a test showed a channel was not incremental. The answer tells you whether they will defend your budget or their retainer.
### 3. "Show me your creative production model."
Paid social in 2026 rewards volume and iteration speed. Static assets and one hero video per quarter will not compete. Ask how many creative variants they ship per month, who produces them, and how fast they turn around a losing concept. Agencies leaning on AI-assisted production should be able to explain their review process, because unreviewed AI output is a brand risk, not a shortcut.
### 4. "How do you report on pipeline, not just leads?"
If your sales cycle is longer than a month, lead volume is a vanity metric. Ask for a sample report. Look for CAC by channel, LTV-to-CAC ratios, and stage-by-stage conversion. Agencies that only report MQLs are optimizing for their own dashboard, not your revenue.
### 5. "What is your position on AI in your own workflow?"
You want specifics, not enthusiasm. Which tasks are automated (briefs, ad variants, reporting pulls) and which stay human (strategy, client communication, final creative review)? Ask how they handle data privacy when using AI tools on client accounts. European clients should also ask where data is processed, given [GDPR exposure](/tech/blog/ai-safety-showdown-what-the-us-china-talks-mean-for-your-data).
### 6. "Who actually does the work?"
The people in the pitch are often not the people on the account. Ask for the names, roles, and time allocations of everyone who will touch your business. Then ask what happens when someone leaves. Agency turnover is the single most common reason retainer relationships sour in year two.
## A quick scoring table
Use this in the room. Score each agency 1 to 5 per row.
| Criterion | What a 5 looks like |
|---|---|
| AEO and search adaptation | Measured AI citation or answer visibility for a client |
| Incrementality and measurement | Runs holdout or geo tests, reports pipeline metrics |
| Creative velocity | Ships 20+ variants monthly with a review process |
| AI workflow transparency | Clear split between automated and human tasks |
| Team continuity | Named staff, documented handover plan |
| Commercial model | Fees tied to outcomes you can verify |
Anything scoring 3 or below is a conversation, not a disqualification. But three or more low scores should end the process.
## What we recommend
Our take, based on what we see working across US, UK, and European accounts:
- For measurement and incrementality work, hire specialists. **Measured** and **Arena** style consultancies exist precisely because most full-service agencies treat testing as an afterthought. If your budget supports it, split measurement from execution so nobody grades their own homework.
- For paid media at scale, look at agencies with genuine creative in-house. **Tinuiti** and **Brainlabs** have the production depth to feed modern ad platforms. Smaller brands should prioritize creative velocity over channel breadth.
- For SEO and AEO, avoid anyone selling "AI SEO" as a standalone product. The work is structural: clean entity signals, citable formatting, and content that answers real questions. **Ahrefs** and **Semrush** both now track AI visibility, so ask which tool the agency uses and why.
- Keep strategy in-house if you have one strong marketing lead. Agencies are better at execution velocity and specialized testing than at owning your positioning. Hire for the gap, not the org chart.
The pattern across all four: pay for verifiable output, not for proximity to trends.
## Red flags in the 2026 pitch room
A few phrases should make you cautious. "We are AI-first" without specifics. "We will figure out attribution later." Any deck that lists 2026 trends without connecting them to your CAC or LTV. And the classic: a proposal that quotes last year's benchmarks for this year's channels. CPMs and conversion rates move. An agency citing stale numbers is either lazy or hoping you will not check.
One more: agencies that cannot name a client who left them, and why. Every good agency has lost business. How they talk about it reveals more than their case studies.
## FAQ
### How long should an agency selection process take?
Six to eight weeks for a mid-market company. Faster than that and you are skipping reference checks or a paid test project. Both matter more than the pitch.
### Should I run a paid pilot before signing a retainer?
Yes, if the budget allows. A four to six week paid test on one channel gives you real data on communication speed, reporting quality, and whether their creative actually performs. Structure it with defined success metrics upfront.
### What if the agency is strong on trends but weak on our industry?
Industry knowledge is learnable in a quarter. Measurement discipline and creative velocity are not. Weight the operating model higher than vertical experience, unless your category has heavy regulatory constraints.
Frequently asked questions
1. "How do you handle answer engine optimization for a client like us?"
A weak answer: "We optimize for featured snippets."
A strong answer names the mechanics. They will talk about structuring cont
Six to eight weeks for a mid-market company. Faster than that and you are skipping reference checks or a paid test project. Both matter more than the pitch.
Should I run a paid pilot before signing a retainer?
Yes, if the budget allows. A four to six week paid test on one channel gives you real data on communication speed, reporting quality, and whether their creative actually performs. Structure it with defined success metrics upfront.
What if the agency is strong on trends but weak on our industry?
Industry knowledge is learnable in a quarter. Measurement discipline and creative velocity are not. Weight the operating model higher than vertical experience, unless your category has heavy regulatory constraints.
DGTG
