I spent $4,800 on AI marketing tools. 3 were worth it.
Twelve months. Seventeen AI marketing tools. Approximately $4,800 in subscriptions, mostly USD-priced, converted to AUD on the day. The honest accounting: I cancelled fourteen of them. Three are still on the card. Here's what each delivered, what each promised, and what made the difference.
Why I'm writing this
Every marketing agency owner I know is being pitched ten AI tools a week. Most of those pitches lead with the same three lines: "saves you hours," "automates your X," and "10x your output." The pitches rarely lead with what the tool actually does, what it costs, or what fails. I wanted to document the year as honestly as I could so the next agency owner has something to read other than a vendor's case study.
Catnapweb is a small agency. Our context shapes what worked and what didn't — a 50-person agency would reach different conclusions. But the categories of failure I'll describe are stable across business sizes: capability overlap, AI adds no incremental value, integration friction, and pricing mismatched to value. Read this as a pattern guide, not a verdict.
What I bought (the full list)
Seventeen tools across five categories. AUD figures are approximate, converted from USD pricing at the rate during my trial window. Several of these have changed pricing tiers since I evaluated them — verify current pricing on each vendor's site before subscribing.
- Content & writing: ChatGPT Team, Claude Pro, Jasper, Copy.ai, Writesonic
- SEO & content optimisation: Semrush, Ahrefs, SurferSEO, Frase, Clearscope
- Meetings & workflow: Otter.ai, Notion AI, Loom AI
- Email & CRM AI: HubSpot AI add-on, Mailchimp Intuit Assist
- Image & visual: Midjourney, Canva Pro
Combined trial period: November 2025 through April 2026. Combined cost across the year: approximately $4,800 AUD. Two tools account for the majority of that figure — premium Ahrefs and Semrush subscriptions together were ~$3,400. The remaining $1,400 spread across the other 15.
The three I kept
The baseline. I use it 30+ times a week — drafting client emails I want to be sharp without being cold, brainstorming article angles, breaking complex topics into outlines, summarising long PDFs, generating first-pass copy I then rewrite. None of these tasks is impressive on its own. But the cumulative time saved over a year is genuinely material — by my estimate, roughly an hour a day of friction removed.
What it's not good at: anything where the output needs to be definitively accurate without verification. AI Overviews-style "what's the current spam policy on this" answers are not reliable. I treat ChatGPT as a writing partner that needs editing, not as a research tool.
Why ChatGPT Team specifically (not the $20 Plus tier): team workspace lets me separate client work from internal work, and the data retention defaults are sensible for an agency handling other people's information. Worth the price gap.
Counter-intuitively, the most important AI marketing tool in 2026 is still mostly a traditional SEO tool. Semrush's AI features (Content Marketing Toolkit, AI keyword grouping, AI-assisted writing scoring) are fine but not why I keep paying. I keep paying for the keyword databases, the backlink crawl, the competitor research, and the position tracking — capabilities no AI tool I trialled actually replaced.
The AI search visibility tracking is the genuinely useful new feature: I can see which of my client sites are being cited in AI Overviews and Perplexity, what queries trigger the citations, and how it shifts over time. That data isn't available anywhere else at this price point.
I trialled Ahrefs alongside it for four months. Ahrefs has better backlink data; Semrush has better AI visibility tracking and better AU-region keyword data. For Catnap's mostly-AU client base, Semrush won. For a global agency, the choice could legitimately go the other way.
The smallest of the three but possibly the highest ROI per dollar. Otter joins every client call, transcribes everything, and produces a structured summary I can paste into a follow-up email. The summaries aren't perfect — they miss nuance about who committed to what — but they catch the action items, the questions raised, and the specific numbers discussed, which is exactly the content I'd otherwise spend 30 minutes reconstructing from memory after each meeting.
The unlock isn't "I have a transcript." The unlock is that I can stay present in the conversation instead of taking notes. That's the bit no marketing-pitch description captures, and it's the bit that actually changed how my client meetings feel.
The fourteen I cancelled (and why)
Grouped by the category of failure rather than the tool, because the patterns are more instructive than any individual review.
Category 1 — Capability overlap with ChatGPT
Category 2 — AI doesn't add value to the underlying task
Category 3 — Pricing mismatched to value at SMB scale
Category 4 — Integration friction
My 2019 paper on machine learning adoption in marketing (IEEE Access, 293 citations) found that tool adoption success is predicted more strongly by organisational maturity than by tool sophistication. The implication for AI marketing tools in 2026: the right question isn't "is this tool good," but "does our process actually use what this tool produces." Most of the tools I cancelled were good tools. They just produced outputs I didn't use, or produced them in ways that didn't fit how I work.
What I'd buy if I were starting over today
If a marketing-agency owner asked me what to subscribe to today with a $400/month budget, my answer would be:
- ChatGPT Team or Claude Pro — pick one, don't pay for both. ~$45/mo
- Semrush or Ahrefs — pick one, don't run them in parallel. ~$190/mo
- Otter.ai or similar — meeting transcription is genuinely transformative. ~$30/mo
- Canva (free or basic) — most agencies don't need the Pro tier. ~$0–15/mo
That's ~$280/month total. The remaining $120/month: don't spend it. Save it for the tool you find in six months that actually fits a specific gap. The single most expensive mistake an agency owner can make in this category is committing to a tool stack before they know which gaps the stack is meant to fill.
Want help auditing your current AI marketing stack?
Our free AI marketing audit looks at the tools you're currently paying for, the capability overlaps, the gaps, and where AI is actually moving the needle for your business — versus where you're paying for branded boilerplate. AU SMBs only.
The bottom line
The three tools I kept aren't impressive. They're a writing assistant, a keyword database, and a meeting transcriber. None of them is the kind of thing a vendor pitches as "transformative AI." That should tell you something. The category of tool worth paying for is the category that does one job well, fits into how you already work, and would be missed if it disappeared tomorrow.
The category of tool to be sceptical of is the category that promises to "10x your marketing output." If a tool's pitch is more impressive than the work product it makes you capable of, you're paying for the pitch.
For an unrelated but adjacent perspective on AI marketing maturity, see what 1,628 academic citations taught me about AI in marketing — the gap between what research says works and what marketing tools promise is wider than most agency owners admit.