Category: AI

Picture your best salesperson cloned, working every search result, available at 3 am on a Tuesday. No commissions. No bad days.

That’s roughly what Google just shipped. They’re calling it the Business Agent, and if you run a retail brand, it’s the most important new tab in Google Merchant Center (GMC) right now.

What the GMC Business Agent Actually Does

When someone searches for your brand or products, they’ll now see a “Chat” option. Click it, and they’re talking to an AI agent that knows your catalog and speaks in your brand’s voice. Previously, when a customer searched for your brand or products, they saw static links and shopping ads. When the Business Agent is available, they will be able to enter a conversation with this AI agent that acts as your Virtual Sales Associate.

It doesn’t just spit out impersonal, templated answers. It uses your Merchant Center product data and your website information to answer specific questions like:

• “Does this rug come in a larger size?”
• “Is this jacket waterproof or just water-resistant?”
• “What accessories go with this camera?”

The GMC Business Agent is not a generic chatbot. It carries your logo, your colors, and the tone you set.

Why This Matters

High-intent shoppers asking specific questions are usually one unanswered question away from leaving. A customer who types “does this fit a king mattress” already wants to buy. They just need someone to tell them yes.

The Business Agent is that someone.

And right now, it’s free. Google is offering this to eligible merchants at no extra cost inside Google Merchant Center, which means the window to get ahead of competitors who haven’t noticed yet is open.

What You Can Control

Google gave merchants more control here than I expected:

Identity: Upload your logo and hex codes. The chat window looks like your site, not Google’s.

Voice & Tone: Pick how your agent speaks. Professional, friendly, casual. This matters more than it sounds.

Welcome Message: This is the first thing customers see. The message “Ask me anything about our spring collection” beats a generic greeting.

Conversation Starters: Include pre-set questions that push shoppers toward your bestsellers.

Support Handoff: Set the rules for when the AI should pass someone to a human agent.

Who’s Eligible Right Now (March 2026)

In order to be eligible, companies must meet the following four criteria:

• US-based accounts (international rollout coming)
• Verified Google Merchant Center account
• At least 50 approved offers in your feed
• Claimed Brand Profile in GMC

The Business Agent Setup Takes About Five Minutes

• Log in to Google Merchant Center
• Go to the Marketing tab
• Click Business Agent
• Click Customize
• Upload your branding and write your welcome message
• Preview it against sample product questions
• Publish

That’s it.

What’s Coming Next: Agentic Checkout

Here’s where it gets interesting. Later this year, Google plans to add Agentic Checkout through something called the Universal Commerce Protocol.

The customer won’t just ask about a product. They’ll be able to say “buy it,” and the agent will handle payment and shipping without them ever leaving the chat window.

No click to your site. No cart. No checkout page. Just a conversation that ends in a purchase.

If you haven’t thought about what that does to your attribution model, you should start.

The Short Version

Your Merchant Center data has been sitting there doing nothing between campaigns. The Business Agent turns it into an active conversation with shoppers. If you’ve got the tab, set it up today. The lift is low, and the downside of waiting is watching a competitor show up in your branded searches with a chat window you don’t have.

AI-generated creative is already changing how advertisers approach visual production. With Google’s Gemini platform rolling out deeper integrations across Google Ads, its built-in image generation model, Nano Banana, gives marketers a practical tool for creating scalable, brand-aligned ad visuals quickly and affordably.

Whether you’re an e-commerce brand looking to improve product presentation or a service-based business focused on generating leads, AI-generated imagery can support nearly every stage of your campaign. The key is knowing how to write prompts that get results. This post covers everything you need to know to get started, including prompt structures, use cases, and real examples from campaigns we’ve tested.

What Is Nano Banana and How It Works

Nano Banana is Google’s native image-generation model built into the Gemini AI platform. It’s been integrated directly into tools like Performance Max, the asset library, and Product Studio. You enter a written prompt into Gemini, and Nano Banana renders an image based on your description. You can guide lighting, product placement, image size, setting, and even edit elements after generation. Need to remove the background, change the scene to a holiday setting, or make the lighting warmer? Just ask.

Be creative, think outside of the box, and embody your inner graphics designer/photographer. The more descriptive you are in defining what you want, the better the system works in recreating your prompts.

Because it’s built for marketers, Nano Banana also supports ad-safe formatting and understands basic design intent, like how to frame a product or keep the brand’s color palette intact.

Nano Banana access is available through two main Gemini tiers (Gemini is integrated into Google Workspace for most business, enterprise, and education plans):

• Free Tier: Good for basic testing and prompt experimentation, but image generation may be limited or unavailable depending on your region.
• Gemini Advanced ($19.99/month): Includes full access to Nano Banana Pro, supports high-resolution outputs, and is best suited for businesses creating visuals regularly.

For consistent use, marketing teams should opt for the Gemini Advanced plan. Product Studio access within Google Ads is still rolling out and may offer limited Nano Banana functionality depending on your account setup.

Why Marketers Should Care About Nano Banana

Nano Banana isn’t just a shortcut to get more visuals. It’s a way to remove creative bottlenecks entirely. Case studies from early adopters show a 20% lift in click-through rates and a 15% decrease in cost-per-lead, demonstrating the tangible impact on marketing efficiency. Here are a few of the ways brands are already using it:

• For E-commerce Brands: If your current product images are plain white background shots, Nano Banana lets you generate realistic lifestyle backgrounds that make your product stand out. You can transform a generic pack shot into a scene with lighting, environment, and props that align with your brand. Need visuals for your website, a carousel ad, a seasonal promotion, or a last-minute social post? You can create it all in-house with a little patience and the right prompt.
• For Service-Based Businesses and Lead Gen: Don’t have a product? No problem. You can use Nano Banana to illustrate your service visually by showing a technician in action, a friendly team photo, a metaphor like a glowing shield for security, or an outcome like a customer smiling in a freshly cleaned home. This helps you build trust and clarify what you offer, without relying on stock photos or custom shoots.
• For Marketing Teams of Any Size: You don’t need a design background to make great visuals. With the right prompts, anyone can create ad-ready images. You can also edit previous generations or quickly create visual variations for A/B testing or seasonal updates. This saves time, cuts costs, and helps smaller teams scale creative output without hiring or outsourcing.

How to Write Prompts That Work

Think of a Nano Banana prompt like a creative brief. The clearer your instructions, the closer the AI gets to your vision. A good prompt describes what the image should show, how it should look, and what kind of emotion or action it should convey.

A simple structure to follow:

Subject | Action | Setting | Style | Lighting | Mood | Optional Details

Focus on:

• Naming the exact object or subject
• Defining the context (where it is, what it’s doing)
• Choosing a style (photorealistic, editorial, flat lay, etc.)
• Explaining the desired light source and mood

Avoid vague language like “beautiful” or “aesthetic.” Use specific visual cues like “soft shadows,” “shallow depth of field,” “overhead camera angle,” or “sunlight from the left.” The AI takes each part literally.

An important thing to remember is that if you don’t deliberately say “create an image,” make sure to select the Nano Banana “Create Image” button at the bottom of the prompt.

Nano Banana Prompt Strategies for E-commerce

For e-commerce advertising, strong visuals can increase click-through rate, build brand trust, and improve conversion. Here’s how to structure prompts based on campaign type or asset need.

• Studio Product Shots: These work best for Shopping feeds, product listings, or clean performance creative.
• Lifestyle Product Scenes: Great for Performance Max ads, Meta ads, hero banners, or social posts where you’re trying to show the product in use.
• Promotional or Seasonal Assets: Quickly update product visuals with holiday themes or new backdrops.
• Multi-Product or Bundled Creatives: Use one prompt to feature multiple SKUs or cross-sell opportunities.

Real Client Example of a Nano banana Prompt

We’ve had the opportunity to work with amazing brands like Jura, known for their premium coffee machines. For this example, I’m using their GIGA 10 Diamond Black as the featured product.

Alongside the prompt below, I’ll also upload Jura’s standard product image – typically a white background studio shot – which helps Nano Banana better understand the product design and preserve accuracy during generation.

Example Prompt: “Photorealistic image of a luxury Jura GIGA 10 Diamond Black coffee machine placed on a high-end kitchen counter with subtle under-cabinet lighting, marble backsplash, early morning sun casting a warm glow from the left, slight steam rising from a nearby cup, and a shallow depth of field.”

After about 30 seconds, this was the generated image:

Prompt Strategies for Lead Gen and Services

If you’re advertising services, software, or anything intangible, image prompts are about evoking emotion, trust, and clarity.

• Human-Focused Trust Visuals: Show real people helping others, using your service, or smiling post-service. This works well in local and B2B ads.

Example Prompt: “A digital marketing strategist sitting at a desk reviewing a Google Ads dashboard on a dual-monitor setup, natural lighting from a nearby window, JumpFly branding subtly on a notebook or mug, clean and modern workspace, focused and confident expression.”

This prompt promotes JumpFly’s Google Ads management services by visually showing the kind of dedicated, data-driven attention a client’s account receives.

• Outcome-Driven Prompts: Display the end result: clean homes, happy clients, streamlined dashboards, or completed work.

Example Prompt: “A Google Ads dashboard showing rising performance graphs on a laptop screen in a bright workspace, with a happy business owner in the background giving a thumbs-up, brand colors softly integrated in the environment.”

• Conceptual Metaphors: For services that are harder to visualize, like data protection or consulting, use metaphors. Shields, bridges, puzzle pieces, and clean paths all communicate value without showing the service directly.

Example Prompt: “A stylized scene of a clean, futuristic workspace where data streams shaped like light trails converge into a glowing, upward-pointing arrow. The arrow is composed of digital elements like search bars, ad charts, and KPI icons, symbolizing campaign growth. Soft gradients of JumpFly blue subtly frame the path, giving the sense of organized momentum and professional clarity.”

How to Scale and Test Efficiently

Once you have a few base prompts that work, you can iterate quickly by changing key variables: the setting, lighting, background, or color treatment.

Save winning prompts. Build a small library of structured formats that align with your campaigns: promos, product launches, testimonials, seasonal refreshes, and remarketing.

You can also test multiple versions side by side. Try a clean product image against a lifestyle one. Add human faces to one version, and remove them from another. Nano Banana gives you the power to test creativity at scale with minimal effort.

Final Takeaways and Things to Remember

Writing effective image prompts takes some trial and error, but it’s already proving to be a valuable skill for modern marketers. Whether you need a clean product photo, a relatable lifestyle visual, or something more conceptual for lead generation, Gemini and Nano Banana give you a flexible way to create those assets quickly.

Start by understanding the structure, test different variations, and fine-tune as you go. With a little creativity and a clear idea of what you want to express, anyone on your team can use this tool to generate visual content that drives results.

OpenAI announced in January 2026 that ads would begin appearing in ChatGPT. This is a big deal. The company that changed how millions of people search for information is now entering the advertising business, and with 800 million weekly users, marketers need to pay attention.

Access is currently limited to select enterprise partners, but that won’t last forever. Here’s what you need to know to be ready when the platform opens more broadly.

How ChatGPT Ads Work

ChatGPT ads appear below the chatbot’s responses, clearly labeled as sponsored content. Unlike traditional Search ads, where users scroll through a list of links, these placements integrate directly into the conversational flow.

Think about it: someone asks ChatGPT for laptop recommendations and sees a contextually relevant ad right after receiving their answer. That’s advertising at the exact moment of decision.

OpenAI has made several commitments around trust and transparency. Advertising won’t influence organic responses. User data and conversations won’t be sold to advertisers. Users can see why they’re being shown specific ads, dismiss irrelevant ones, and turn off personalization entirely.

Who Sees Ads

Only free users and ChatGPT Go subscribers ($8/month) will see ads initially. Premium tiers, including Plus ($20/month), Pro ($200/month), Business, and Enterprise, remain ad-free. Users under 18 won’t see ads, and OpenAI is avoiding placements near sensitive topics like politics, health, and mental health.

This mirrors what we’ve seen from streaming platforms: a free or low-cost option supported by advertising (think Google Shopping ads at the very beginning, AKA Froogle), with premium ad-free experiences for those willing to pay. For advertisers, this means the initial audience skews toward price-conscious users who haven’t committed to a paid subscription.

Premium Pricing, Limited Measurement

OpenAI is targeting a Cost-per-thousand impressions (CPM) of around $60.

For context: Google Display Network ads average around $3 CPM. Google Search ads sit closer to $38 CPM. Premium connected TV inventory from Hulu and Netflix’s ad tier typically commands $40 to 65 CPM. OpenAI is pricing at the very top of the digital advertising spectrum.

But why? ChatGPT captures users at moments of genuine intent. When someone asks about running shoes for flat feet, they’re expressing unfiltered purchase consideration before narrowing their choices. That’s the high-intent moment advertisers have always coveted.

The trade-off is measurement. Unlike Google and Meta’s sophisticated attribution systems built over two decades, OpenAI is launching with basics: total impressions and total clicks. No granular conversion tracking. No demographic insights. No purchase attribution.

OpenAI says more detailed data may come over time, but for now, performance marketers face real challenges justifying spend without visibility into what’s actually driving results.

The Early Advertiser Program

OpenAI is reaching out to advertisers directly through its enterprise partnerships team, not through traditional agency channels. The initial trial targets companies with significant spending commitments over several weeks, with ads launching in early February.

This enterprise-first approach differs significantly from how Google and Meta built their ad businesses on the backs of small businesses using self-serve tools. There’s no self-serve buying interface yet, and OpenAI isn’t working with ad tech firms to manage placements.

For agencies and smaller advertisers, this means watching from the sidelines initially, but that’s exactly when smart marketers start doing their homework.

What This Means for Your Strategy

ChatGPT’s entry into advertising signals a shift in how consumers discover products. As more people turn to conversational AI instead of traditional search, advertising dollars will follow. The implications depend on your current approach:

• Search advertisers should view ChatGPT as a new high-intent channel that could capture queries currently going to Google. The conversational format suggests users may be further along in their decision-making process. However, the lack of keyword-level targeting and bidding controls makes optimization difficult compared to Search campaigns, where we’ve spent years refining strategies.
• Brand advertisers may find the premium pricing and brand-safe environment appealing for awareness campaigns. Being associated with ChatGPT’s helpful, authoritative persona could carry positive brand associations, particularly for companies looking to align with innovation.
• Performance marketers face the biggest hurdle with limited attribution data. Without visibility into which ads drove conversions, campaign optimization becomes guesswork. Early adopters should approach with clear testing frameworks and realistic expectations.

Our Take on ChatGPT Advertising

At JumpFly, we see ChatGPT ads as one of the most exciting developments in digital advertising in years.

With hundreds of millions of active users turning to ChatGPT for recommendations, research, and discovery, OpenAI has built an audience that most platforms would envy. The opportunity to reach consumers in a high-intent, conversational environment, right at the moment they’re actively seeking solutions, is significant.

We’re eager to get our hands on this platform. Understanding how to create compelling ads for a conversational interface, learning what optimization levers exist, and discovering how to drive measurable success for our clients on an entirely new channel is exactly the kind of challenge our team thrives on. New platforms reward early adopters who invest the time to learn their nuances, and we intend to be ready.

Yes, measurement is limited today, and pricing is premium. Like any responsible agency, we’ll approach testing with appropriate caution and clear expectations. But limited reporting at launch is the norm for emerging platforms. Meta and Google weren’t built in a day either.

What matters is the potential. ChatGPT ads have the potential to become a core component of many brands’ digital advertising strategies.

Our recommendation: lean in. Work with partners committed to understanding this new frontier. Test early to build institutional knowledge. Position your brand to capitalize as the platform matures. Conversational AI advertising isn’t a passing trend; it’s where the industry is heading. The brands that engage now will have a meaningful advantage as ChatGPT’s ad product evolves.

Looking Ahead

OpenAI’s move into advertising validates that AI-powered interfaces have become significant enough to warrant their own advertising ecosystems. The next several months will be telling:

How do early advertisers perform? How do users react to ads in their conversations? Will OpenAI develop the measurement and targeting capabilities that performance marketers need?

These questions will determine whether ChatGPT becomes a meaningful part of the media mix or remains a niche experiment. What we know for certain is that the way people discover information is changing, and advertising will adapt to follow.

ChatGPT ads are just the beginning of what promises to be a fundamental shift in how brands connect with consumers in an AI-first world.

The conversation around AI in paid Search has shifted. It’s no longer about whether you’re using these tools; it’s about how much control you’re willing to hand over. 

We’ve all seen the headlines promising that AI will replace the creative department. We’ve watched how easy it is to generate “limitless” ad content. But now that the dust has settled, a different reality has emerged: the biggest challenge isn’t making AI smarter. The real issue is figuring out exactly when a human needs to step in and stop the machine from being generic, or even from being downright wrong. 

To win today, you must become a Cyborg Architect. Let the machine do the heavy lifting and the grunt work, but you stay firmly in the driver’s seat.

AI is an incredible concept engine. It can spit out 50 search headlines in the time it takes you to open a new Google Doc. But letting the AI hit “Publish” on its own is a dangerous game. When we treat AI as a vending machine rather than a brainstorming partner, we get a heap of AI Slop, that soulless, blah ad copy that searchers have already started to tune out.

The competitive advantage no longer goes to the person with the best prompts. It goes to the person who knows how to curate, polish, and stress-test the results AI produces.

A Reality Check on “Hallucinations”

In the early rush to use AI for creating ads, people started using LLMs to pull final ad copy directly from website URLs. The promise was speed, but the reality was a mess of “hallucinations,” where AI confidently produced things that were just flat-out wrong.

I saw this firsthand recently. I asked ChatGPT to generate headlines for a client’s Responsive Search Ads (RSAs). The copy looked professional and polished, claiming the company had been in business since 1973.

Something felt off. I checked the site; the business was actually founded in 1978. When I called the AI out, it doubled down, insisting the date was on the landing page. Only when I asked for the specific sentence did it cave: “Apologies, you are correct!” This isn’t an isolated glitch. For another client, the AI invented a “Free Shipping Over $65” offer. The problem? That client never offers free shipping. If those ads had gone live, we would have burned through the budget while destroying the brand’s reputation. Around our office, we have a mantra: “Trust, but verify.”

This is why the Cyborg model is so important. An AI can process data, but it doesn’t always get the truth right. Without a human ‘pilot’ to perform a manual override, speed simply becomes a faster way to fail.

Escaping the “Sea of Sameness”

Beyond the risk of lying to customers, AI-only copy leads to a sea of sameness. Think of AI as a giant blender of the entire internet. It’s just spitting back the mathematical average of what everyone else has already written. It’s designed to be ‘normal,’ but in PPC, ‘normal’ is just another word for invisible.

You’ll get the same tired tropes: “Elevate your run,” “Unlock performance,” “Solutions for you.” In fact, our team has grown to genuinely loathe the word “elevate.” AI seems to be obsessed with it. In a world where you only have seconds to grab attention, being average is a death sentence for your ROAS.

The Solution: The “Sandwich” Workflow

Instead of viewing AI as an “Author,” view it as a “Research Assistant.” We use a Sandwich Workflow to balance machine speed with human judgment.

1. The Bottom Bun: Human Architecture

You don’t start with the AI. You start with the human strategist who sets the gears in motion. This is where you define the goals, the brand voice, whatever guardrails that need to be in place, and the actual strategy. 

2. The Filling: The AI Rapid Volume

This is where you let the machine off the leash. As humans, we’re wired to stop once we hit a “good enough” idea so we can move on to the next task. AI doesn’t get tired. It’s a limitless engine of unfiltered variations. It can spit out 100 different angles you may have never thought of in mere seconds. While 90 of them might not be what you’re looking for, it does drastically increase the odds of finding that one creative outlier that piques your interest.

3. The Top Bun: Human Curation & Deployment

This is where we bring order to the chaos. This isn’t just “editing”; it’s about logic, empathy, and a “vibe check” that AI simply can’t simulate. The human’s job is to sift through the noise, filter out the hallucinations, and ensure the final ad actually resonates with a living, breathing person on the other side of the screen.

Sandwich Layer	Mode	Role	The Cyborg Synergy
Bottom Bun	Architecture	Human	Sets the goals, brand voice, and guardrails.
The Filling	Rapid Volume	AI	Generates 100+ variations and creative angles.
Top Bun	Curation & Execution	Human	Trust but verify. Filters for truth and “soul.” Hacks away the generic fluff to create high-performance ads that actually sound like a person wrote them.

The Rise of the Cyborg Architect

The future of high-performance paid Search isn’t a choice between human intuition and machine speed. It’s the synthesis of both.

We are entering the era of the Cyborg Architect. This isn’t about handing the keys to a bot and crossing our fingers; it’s about using AI as a power-up. Let the machine handle the grunt work of churning through thousands of ideas. Your job is to provide the judgment, the ethics, and the soul of the brand, the “stuff” a machine can’t even begin to replicate.

The most successful brands won’t be the ones using the most AI. They’ll be the ones who know exactly where the machine ends and the human begins.