A Perth retailer sat through four agency pitches last quarter. Three opened with an AI slide. Bidding automation, content generation, a dashboard with a robot icon on it. The retailer picked the fourth agency — the one that spent ten minutes explaining which decisions it still makes by hand and why.
That’s the shift worth understanding before you hire anyone in 2026. Salesforce’s State of Marketing 2026 puts generative AI use at 87% of marketers, in at least one workflow, up from 51% just two years earlier. McKinsey’s global survey found 82% of organisations now run AI somewhere in the business, and marketing is the single most common place it shows up.
Once a capability is that common, it stops being a selling point. Everyone has it. What separates the agencies worth paying from the ones coasting on a buzzword is narrower and harder to fake: what job is the AI actually doing, and who’s still making the call when it matters.
A few Perth teams already answer that question well. One marketing agency Perth breakdown lays the split out cleanly — AI absorbing the volume work, strategists keeping the judgment calls. Worth listening for that exact division in any pitch you sit through.
Google Ads: where automation quietly overspends
Paid search is the easiest place for AI to look productive while burning budget in the wrong direction.
Google keeps folding more of Ads into AI Max — generated copy, audience signals pulled from behaviour data, bids adjusted in real time off conversion history. Left alone, that system will spend efficiently toward whatever goal it was given. If the goal was set loosely, it optimises toward the wrong customer just as confidently as the right one.
Find out which levers are automated and which still require manual control. Then dig into how wasted spend is identified before it compounds over a month, and whether the agency measures lead quality rather than quietly treating “more clicks” as a substitute for “more revenue.” A lot of agencies struggle to answer that last question clearly. That’s usually the tell.
SEO has split into two jobs, not one
Ranking on page one still counts. It’s just not the only game running anymore.
People now get answers straight from Google AI Overviews, ChatGPT, Gemini, Perplexity, Copilot — no click required. HubSpot’s 2026 State of Marketing lists updating SEO for this shift as one of the top five priorities marketers named, at 40.6%.
The mechanics an agency should be handling:
- Search intent — matching the actual question, not the keyword string
- Entity clarity — making the brand a recognisable, citable “thing” rather than a page full of text
- Structured content — schema, clean headings, answers stated early
- Topic depth over one-off posts
- Earned mentions from sources an AI system would trust citing
Google’s own guidance hasn’t changed much: write for people, not the algorithm. What’s different is that content built this way now has a shot at showing up in two places instead of one.
The question almost nobody asks yet
Try this one on a pitch and watch what happens: how would you get us mentioned inside an AI-generated answer, not just ranked on a page?
Somebody typing “best [service] Perth” into ChatGPT gets a handful of names back, not ten blue links to scroll through. Getting onto that shortlist runs on different signals than classic SEO — brand authority, third-party mentions, reviews, comparison writeups, how often competitors already get cited by name.
Most agencies haven’t built a real answer to this yet. That’s not necessarily disqualifying on its own. What is disqualifying is an agency that hasn’t even noticed the question exists.
Content: the easy part is the trap
Generative AI makes producing content nearly free. That’s exactly where the trouble starts.
A hundred generic articles by Friday afternoon, sure. Content that sounds like it came from someone who actually knows the industry, earns a bit of trust, and moves someone toward buying — that still takes work AI can’t shortcut on its own.
The useful split: AI handles keyword research, competitor scans, first-draft structure, repurposing one piece into five formats. A person still edits for accuracy, tone, and whether there’s an actual point being made. Skip that last step and every article ends up sounding like every other AI-written article on the internet — which, ironically, makes AI use a net negative for the client.
Personalisation and what happens after the form gets submitted
Two things get skipped in most agency conversations: whether messaging actually changes by audience, and what happens the moment someone enquires.
AI-driven personalisation means a plumber’s landing page can read differently for someone searching “emergency” versus someone comparing quotes for a renovation — instead of one script for everyone. HubSpot’s 2026 data has AI-personalised content as the single top trend marketers named this year, at 48.57%.
Follow-up matters at least as much, arguably more. A form sitting unanswered for six hours has already lost real value. Systems now respond instantly, qualify, book a call, or trigger a sequence without anyone touching a keyboard. So ask plainly: does this agency only generate leads, or does it own what happens next? Because a lead nobody calls back is ad spend with nothing to show for it — and that’s on the agency as much as the client.
Creative production is getting faster, not more automatic
Ad images, video concepts, storyboards, copy variations — generative tools now let a team test far more directions than was practical a few years back. Clico’s approach to turning prompts into finished images and video is worth a look for how that pipeline runs day to day: more volume of raw concepts, same amount of human editorial judgment deciding which ones ship.
That’s the correct version of this. Speed of testing goes up. The decision about what actually gets published still sits with a person who understands the brand.
Strategy doesn’t come out of a prompt box
AI writes fast headlines, fast ad copy, fast reports. What it doesn’t know is your margins, your actual competitors, your reputation in the market, or what a genuinely profitable customer looks like for your specific business.
Somebody still has to decide who to target, why a customer should pick this business over the one next door, and which channel gets more budget next quarter versus which one gets cut entirely. Those aren’t prompts. They’re commercial bets, and they require someone who understands the business well enough to lose the argument gracefully when the data disagrees with their gut.
The agencies still standing at the end of 2026 will be the ones that paired experienced people with AI — not the ones that tried swapping the people out for it.
A comparison sheet worth bringing to the pitch
| Area | Question to ask |
| SEO | How are you adapting for AI search? |
| Content | How do you prevent generic AI output? |
| Google Ads | Which decisions are automated vs. manual? |
| Analytics | How do you turn data into action, not just charts? |
| Lead handling | Can you automate follow-up, not just capture? |
| Personalisation | Does messaging actually change by audience? |
| Creative | How do you use generative image and video tools? |
| AI Search | Can you track our brand inside AI-generated answers? |
How specific the answer gets tells you more than any slide claiming the agency is “AI-powered.”
What tends to be a red flag
Watch for an agency that:
- Calls itself an “AI agency” but stumbles when asked what that actually means day to day
- Publishes large volumes of AI content nobody edited
- Promises AI can guarantee rankings — nobody can guarantee that, AI or not
- Automates ad spend without anyone watching profitability week to week
- Can’t explain, in plain terms, how customer data feeds whatever AI tools it uses
None of these are rare. They’re the most common ways “AI-powered” gets used as a sales line instead of an actual working method.
The question that actually matters
Marketing skill, the right technology, and genuine commercial understanding — miss any one of the three and results suffer, no matter how much AI sits underneath.
So don’t ask whether an agency uses AI in 2026. Ask what it’s actually improving. Faster opportunity detection. Less wasted spend. Quicker follow-up. Sharper personalisation. A shot at showing up inside an AI-generated answer instead of buried under one. Those are outcomes worth paying for. A tools list on a homepage isn’t.
Related: How AI, AR, and Social Media Are Rewriting Fashion and Beauty Shopping
