George Little came into marketing through architecture, spent years in business-to-business graphic design, and now runs BrandZap, a design consultancy for B2B and SaaS companies that sits deliberately between the solo freelancer and the full agency. On this episode of The Unscripted SEO Interview he and Jeremy Rivera get into buying committees, self-hosted language models, and why AI gradients are the clip art of 2025.
Write for one person, arm them for the room
George’s most counterintuitive advice is that in B2B — where a purchase touches a whole team — you should write copy for exactly one person. For BrandZap that person is the product marketer or marketing manager, because they are the one doing the searching and the outreach.
“The way I’m communicating it is speaking in a language that that person inherently knows and uses. And that actually creates a connection that’s like, this person knows exactly what my problem is.” — George Little
The job doesn’t end at that connection. The site and the collateral also have to hand that person the tools to make the case internally — to the CFO who controls the budget, to the stakeholders who will live with the decision. You are not converting a visitor. You are equipping an internal champion. His website posture follows from it: “let’s get on a consultation and tune this correctly,” not “sign up right now.”
Three buckets for AI — and the third one is the business model
George sorts his AI use into three buckets. First, filling knowledge gaps: reading a contract, working in a domain he has no base knowledge in. Second, compressing his own workflows — meta descriptions, image generation — so he becomes the reviewer instead of the producer. Third, and most strategically, building AI-powered tools for clients that execute on the brand standards he has already defined.
That third bucket is a retention model. Clients self-serve inside guardrails he set, which keeps him embedded in their process without putting him on every deliverable.
“I’m much more open to being like, I built these 10 pages with AI, but they’re all based on my standards and the thing that I taught it. It’s more of a partner than I thought it would be.” — George Little
Disclosure flipped from liability to pitch
Two years ago, admitting AI-assisted work felt risky. George says the dynamic has fully reversed. When a client used Claude to build a deck out of his brand presentation, he didn’t object — he gave them better prompting instructions. The positioning moved from “I did this by hand” to “I’m the person who teaches the AI how your brand works.”
“I’m not presenting that like, do I hire the human or do I secretly use AI? It’s more like, no, this is a consultant who understands AI and probably will help me use it better.” — George Little
The human-bot sandwich
I floated the framing I’ve been using lately — the human-bot sandwich, where a person asks Claude, Claude asks Google, and the answer comes back through layers of machine before it reaches a human. George’s read was grounded.
“That LLM might do the searching, but a human still has to make the decision, sign the contract, and agree.” — George Little
His site architecture advice follows: build pages Google and the LLMs can parse, but design them for the haphazard, multi-device, interrupt-driven way people actually browse. Don’t bet an entire content strategy on one platform’s current rules, because those rules will change.
Stop AI-washing your feature pages
The sharpest take of the hour came out of what George sees across his B2B SaaS client base: companies bolting AI branding onto every feature without saying what the AI does.
“Stop putting AI gradients on every feature you have. Be specific about it. What’s it actually doing? What action is it going to actually help your users take — as opposed to, it uses GPUs?” — George Little
Vague AI claims build a credibility problem, and they don’t earn citations either. The tools with long-term durability will be the ones that describe a specific action for a specific role.
Self-hosted models, because consistency beats capability
One of the more technically interesting turns: George runs open-source models locally on Mac Minis through Open Web UI for repetitive production tasks like meta description generation, specifically because the commercial APIs keep changing behavior underneath him.
“The output is much more tunable and consistent than you get with any given GPT. When I download a model and have it set up to do the thing I want, it’s not getting any updates from any overlay.” — George Little
This is the same thread running through several recent episodes. Whether it’s a defined skill, a self-hosted model, or a tightly configured API call, controlling the model’s variability is what turns a demo into production output. Variability is a feature when you’re exploring and a bug when you need the same result twice.
Where is the next free channel?
George’s question for the next guest: where is the next genuinely open marketing channel? LinkedIn used to be it — free reach, real engagement, community — and now it’s business TikTok. Paid search is priced out. Reddit is a minefield. We landed on moderated Slack communities as the closest current answer, with Skool-style communities emerging behind them. Neither of us had a confident answer, which is probably the honest state of things.
His bottom line for consultants heading into next year: a good consultant should be very aware that the biggest distraction for your clients is AI trying to do the thing you are also trying to do.
Connect with George Little
George runs BrandZap — B2B branding and Webflow websites for growth-stage SaaS and tech companies — at brandzap.xyz, and he’s on LinkedIn. If you Google “George Little AI” he currently outranks a secretary of state and a fictional mouse, which he seems quietly proud of.
For more conversations like this, explore the topical hubs at Unscripted SEO or try the keyword research tools at SEO Arcade.
