Everyone has opinions about AI tools. Here are mine, based on actually using them every day across video production, creative AI, and GEO consultancy.
This isn’t a “top 10 AI tools” listicle. It’s a working list — the tools that have stuck around in my workflow because they genuinely save me time and produce results I’m happy to put my name on. Some of these I use daily. Others I reach for on specific projects. A few I’ve stopped using entirely.
The Daily Drivers
ChatGPT and Claude are the two I open every morning. They serve different purposes, and I’ve stopped trying to make one do everything.
Claude is my go-to for longer-form work. Content strategy documents, detailed research briefs, analysing competitor positioning — anything that needs sustained reasoning over multiple paragraphs. When I’m building out a GEO content plan for a client, Claude handles the complexity without losing the thread halfway through.
ChatGPT is faster for quick iterations. Brainstorming session titles, rewriting a paragraph three different ways, generating structured data snippets. The conversational back-and-forth feels more natural when I’m exploring ideas rather than executing on a plan.
I use both every single day, and the honest answer to “which is better?” is: it depends on what I’m doing in the next 20 minutes.
Visual and Creative AI
For visual work, Midjourney remains the strongest option for concept art and storyboarding. When I’m pitching a video concept to a client, being able to generate mood boards and visual references in minutes rather than hours has genuinely changed how I approach pre-production. The quality is high enough that clients immediately understand the creative direction.
DALL-E fills a different gap — quick mockups, social media assets, and situations where I need something specific and functional rather than artistic. It’s better at following precise instructions, even if the output lacks Midjourney’s aesthetic punch.
The key distinction: Midjourney for “show me something beautiful that captures this feeling” and DALL-E for “give me exactly this layout with these elements.” Both have a place, and I’ve stopped trying to force either into the other’s role.
Video Generation
This is the space that’s moving fastest and where I’m most cautious about overpromising.
Runway has become genuinely useful for specific production tasks. Short motion graphics sequences, animated backgrounds, and visual effects prototyping — these are areas where AI generation saves real production time. For client pitches, being able to show a rough motion concept before committing to a full production budget is valuable.
Pika I use for different kinds of experiments. It handles certain stylistic transformations well, and for social content where the bar is “engaging” rather than “broadcast quality,” it delivers.
But here’s the honest take: AI video generation is not replacing production-quality footage. Not yet, and not soon. What it’s doing is making pre-production faster, concepts more tangible, and pitches more convincing. That’s genuinely useful. Anyone telling you it replaces a proper shoot is either selling something or hasn’t tried matching AI output to broadcast standards.
GEO and Search
This is where AI tools become less about generation and more about understanding. Generative Engine Optimisation is fundamentally about knowing how AI search engines parse, evaluate, and cite content.
I use Claude and ChatGPT as research tools here — testing how they interpret different content structures, analysing which entities they associate with specific topics, and reverse-engineering the citation patterns that get content surfaced in AI-generated answers.
Beyond the LLMs themselves, I work with structured data tooling for schema markup, entity analysis tools for knowledge graph positioning, and content audit frameworks that evaluate pages through an AI-readability lens rather than just traditional SEO metrics.
The shift from SEO to GEO isn’t about abandoning what works. It’s about understanding that the same content now needs to serve two audiences: human readers and AI systems that summarise and cite on their behalf.
What I’ve Stopped Using
A few tools didn’t survive the hype cycle in my workflow.
Most “AI writing assistants” that promise to write entire blog posts or marketing copy. The output reads like exactly what it is — generic, pattern-matched text with no actual perspective. For someone who needs to sound like themselves (which is everyone doing thought leadership), these tools create more editing work than they save.
Several image upscaling and enhancement tools that promised professional results but consistently introduced artifacts that any trained eye would spot. The technology will get there, but for client work, “almost good enough” isn’t good enough.
And a handful of “AI-powered” tools that turned out to be thin wrappers around API calls with a markup. If I can get the same result by prompting Claude directly, I don’t need a middleman charging a subscription.
The Throughline
The tools I keep using share three qualities: they save me measurable time on real projects, the output quality meets professional standards, and they fit into how I already work rather than demanding I restructure my process around them.
AI tools are creative accelerators. The best ones make good work faster. They don’t make bad ideas good, they don’t replace the judgment that comes from years of production experience, and they definitely don’t eliminate the need to understand your client’s actual problem.
The hype cycle is real, but so are the genuine productivity gains. The trick is knowing which is which — and that only comes from using them on real work, not reading about them on Twitter.