How Hubspot Tested Google Gemini Image Generation

The latest experiment from Hubspot with Google Gemini shows marketers what is possible with AI image generation and where the limits still are.

This guide breaks down the original demo, what was tested, and how you can safely apply similar techniques in your own marketing workflows.

What the Hubspot Google Gemini Demo Tried to Prove

The original Hubspot Google Gemini image demo explored whether the model could understand and visually represent complex, real‑world scenarios.

In simple terms, the test asked: can an AI tool produce on‑brand, accurate images from detailed prompts that resemble what marketers need for campaigns, ads, and social content?

Core goals of the Hubspot experiment

• Check how well Google Gemini follows detailed instructions.
• See if AI can image realistic workplace and lifestyle scenes.
• Evaluate how safe and unbiased the outputs appear.
• Understand where human editing is still absolutely required.

Main Prompts Used in the Hubspot Gemini Test

The demo relied on several multi‑step prompts that mirror the types of briefs marketing teams send to designers or freelancers.

Example marketing prompt structure

The Hubspot team used prompts with elements like:

• A clear subject (for example, a professional at work).
• A specific setting (office, remote workspace, conference).
• Emotional tone (collaborative, focused, friendly).
• Brand‑style notes (color palette, modern look, clean layout).
• Usage context (blog header image, social thumbnail, landing page hero).

By combining these details, the test simulated how real marketing teams might describe creative needs to AI.

How Hubspot evaluated the AI images

Each output from Gemini was reviewed along four basic criteria:

1. Relevance: Did the image match the written prompt?
2. Quality: Was the image sharp, coherent, and visually usable?
3. Diversity: Did it portray inclusive and varied characters and settings?
4. Brand fit: Could a similar image fit into a Hubspot style layout or modern B2B blog?

What Marketers Can Learn from the Hubspot Demo

The results of the experiment are useful beyond the Hubspot blog. They highlight practical lessons for anyone considering AI imagery for content and campaigns.

1. AI is strong at concept exploration

Gemini handled idea exploration well. When prompts were descriptive, the model produced several different visual takes on the same brief.

For marketers, that means AI can help you:

• Brainstorm early‑stage visual directions.
• Test different composition ideas quickly.
• Create low‑stakes mockups before investing in custom design.

2. Human review remains non‑negotiable

The Hubspot team still had to verify outputs for realism, bias, and accuracy. Small details like hands, text inside images, or background objects could look off or inconsistent.

In production workflows, that translates into clear rules:

• Always have a human approve any AI visual before publishing.
• Do not rely on AI images for legal, medical, or safety‑critical content.
• Double‑check brand consistency, especially colors, logos, and typography.

3. Prompt clarity drives better results

When the Hubspot prompts were more specific, images generally aligned more closely to what a marketer would expect. Vague prompts produced generic or mismatched images.

Well‑structured prompts tend to include:

• Who is in the scene.
• What they are doing.
• Where they are located.
• How the scene should feel.
• Why the image will be used.

Step‑by‑Step: Recreating a Hubspot‑Style Gemini Test

You can recreate a simplified version of the Hubspot test inside your own organization using the following process.

Step 1: Define your marketing scenarios

List 3–5 realistic scenarios where you would normally brief a designer. For example:

• Blog hero image for a product launch article.
• LinkedIn post graphic announcing new data.
• Email header image for a webinar invitation.

Step 2: Turn each scenario into a detailed prompt

Write a prompt for each scenario using the structure showcased in the Hubspot demo:

• Subject and role (“a marketer presenting results”).
• Setting (“modern open office”).
• Angle or composition (“medium shot, centered subject”).
• Style (“clean, bright, tech‑savvy, minimal clutter”).
• Output type (“wide banner image suited for a blog header”).

Step 3: Generate multiple variations

Request several versions of each prompt from Gemini or another model. Keep track of which settings (resolution, style, seed, or version) you use so you can compare.

Step 4: Score your outputs like Hubspot

Create a simple scorecard based on the criteria from the Hubspot article:

• 1–5 for relevance to the prompt.
• 1–5 for visual quality.
• 1–5 for inclusivity and diversity.
• 1–5 for brand fit.

Invite stakeholders from marketing, design, and content to rate images independently, then compare scores.

Step 5: Document safe‑use rules

Use your findings, plus the lessons from the Hubspot demo, to create internal guidance such as:

• When AI images may be used (for example, blog thumbnails, internal decks).
• Where you still require custom photography or illustration.
• Review and approval steps before anything goes live.

Best Practices Inspired by the Hubspot Gemini Test

Applying the Hubspot insights to your own strategy will help you balance creativity with responsibility.

Combine AI images with design tools

Instead of publishing raw AI exports, route them through your design stack:

• Add brand colors, fonts, and logos.
• Crop or adjust layout for different channels.
• Overlay text and CTAs using consistent templates.

Be transparent with your audience

The Hubspot discussion around AI imagery emphasizes clarity and trust. Consider:

• Disclosing when hero images are AI‑assisted.
• Avoiding AI visuals in sensitive or emotionally charged topics.
• Using alt text that accurately describes the image content.

Continuously monitor AI model changes

As Google Gemini and similar tools evolve, their capabilities and limitations will change. Re‑running a Hubspot‑style test once or twice per year helps you:

• Catch quality improvements you can leverage.
• Spot new failure patterns or biases early.
• Update your internal guidelines before issues scale.

Where to Go Next After the Hubspot Gemini Demo

If you want to apply structured testing and AI governance beyond what the Hubspot example covers, you can explore specialized consulting resources such as Consultevo for broader AI strategy support.

For the full breakdown of prompts, screenshots, and commentary, read the original Hubspot Google Gemini image demo. Use it as a living reference while you design your own AI‑ready content processes.