Google Ads’ New Gemini AI Tool Tested: Is It Actually Any Good?
Google is investing heavily in artificial intelligence, and one of its biggest recent announcements is the integration of Gemini — Google’s most advanced AI large language model — directly into the Google Ads platform.
According to Google, this AI integration is designed to help advertisers create campaigns faster, simplify onboarding, and generate ad assets at scale. On paper, it sounds incredibly promising. If AI can build effective campaigns automatically, it could reduce the barrier to entry for businesses wanting to advertise online.
But does it actually work?
In this article, we’ll break down a real-world test of Google Ads’ new Gemini-powered asset generation system, explore what it gets right, where it fails badly, and whether advertisers should rely on it yet.
What Is Google Ads’ Gemini AI Tool?
The new Gemini AI functionality inside Google Ads is designed to assist advertisers during campaign setup by automatically generating:
- Keywords
- Headlines
- Descriptions
- Images
- Site links
- Other campaign assets
The goal is simple: reduce the amount of manual work required to launch campaigns.
Google positions this as a smarter, faster way to create campaigns, especially for newer advertisers who may not fully understand keyword research or ad copywriting.
To test the system properly, a real plumbing company website from the UK was used during the setup of a standard Google Search campaign.
Step 1: Website Scanning – A Surprisingly Strong Start
The first step in the process involved entering a website URL into Google Ads.
Immediately, Gemini scanned the website and extracted key business information such as:
- Plumbing services
- Heating
- Leak repairs
- Bathroom installations
- HVAC services
- Electrical work
Interestingly, the AI accurately identified the business offerings simply by crawling the website content.
From a usability standpoint, this part worked well.
For new advertisers, automatically understanding the business type and services could save a considerable amount of setup time.
Verdict on Website Scanning
Pass
The AI correctly identified the business and services with minimal effort from the advertiser.
Step 2: AI Keyword Generation – A Major Disappointment
After scanning the website, Gemini moved on to keyword generation.
This is where problems started to appear.
Despite Google having access to one of the world’s most powerful keyword databases through the Google Keyword Planner, the AI generated only five keywords.
Just five.
For a large plumbing company offering multiple services, this is nowhere near enough to build a scalable campaign.
Why This Is a Big Problem
Effective Google Ads campaigns rely heavily on comprehensive keyword research.
Advertisers typically need:
- Core service keywords
- Location-based keywords
- Long-tail variations
- Emergency service searches
- Commercial intent terms
- Negative keywords
Generating only five generic keywords is simply inadequate for real campaign management.
Even more concerning, when prompted to generate additional keywords, the AI refused.
Instead, it responded with:
“I am unable to help you with keywords at the moment.”
That response alone highlights how limited the current system still is.
Verdict on Keyword Generation
Fail
Advertisers should absolutely continue using the Google Keyword Planner and manual research rather than relying on Gemini for keyword discovery.
Step 3: AI-Generated Headlines – Better Than Expected
Next came the ad headlines.
Gemini generated 15 headlines automatically, including variations such as:
- Heating and plumbing services
- Trusted for 40 years
- 24/7 availability
- Emergency electrical services
- Plumbing and electrics
Surprisingly, many of these headlines were reasonably good.
They were relevant to the business and aligned with Google Ads best practices.
However, there were still clear weaknesses.
The Main Issue With the Headlines
Most headlines were too short and failed to fully utilise the available character space.
For example:
- “Heating and plumbing services”
could easily be expanded into:
- “Trusted Heating and Plumbing Services in London”
to improve relevance and click-through rates.
The AI also lacked stronger calls-to-action and more compelling selling points.
Ad Strength vs Actual Quality
An important point raised during the test was the distinction between:
- Ad Strength
- Real-world ad performance
Google automatically rated the ads as having “Excellent” Ad Strength.
But Ad Strength simply measures how closely ads follow Google’s formatting recommendations — not whether they will actually convert customers.
Because Google generated the ads itself, achieving an excellent score was expected.
That does not mean the ads are truly effective.
Verdict on Headlines
Partial Pass
The headlines are a decent starting point, but they still require significant human improvement.
Step 4: AI Descriptions – Generic and Weak
The AI generated only two descriptions, despite Google Ads allowing up to four.
That alone is a limitation.
Worse still, the descriptions themselves were underwhelming.
Instead of focusing on:
- Customer pain points
- Benefits
- Urgency
- Unique selling propositions
the AI largely produced generic “about us” style content.
This is a common mistake inexperienced advertisers make.
Rather than persuading users to take action, the descriptions simply described the company.
Why This Matters
High-performing ad copy should focus on:
- Solving customer problems
- Building trust
- Encouraging action
- Highlighting competitive advantages
The AI-generated descriptions failed to achieve this effectively.
Verdict on Descriptions
Fail
The descriptions were too generic and incomplete.
Step 5: AI Images – Not Really AI
The system also generated image assets.
At first glance, this looked promising.
However, upon inspection, the images were simply pulled directly from the company website rather than being AI-generated.
Examples included:
- Existing branded images
- Company vans
- Staff photos
While these images may still be useful, they are not examples of advanced AI creativity.
Essentially, Gemini was functioning more like a website scraper than a true creative AI system.
The Problem With Website-Based Images
Many service businesses do not have strong visual content on their websites.
Effective advertising often requires:
- Custom-designed creatives
- Service-specific visuals
- Promotional graphics
- Strong branding
Simply recycling website images is not always enough.
Verdict on Images
Pass — but not because of AI
The feature works, but it is far less impressive than Google’s marketing suggests.
Step 6: Site Links – A Serious Context Failure
The AI also generated site links automatically.
Unfortunately, one of the suggested links was:
- Career Opportunities
This highlights a major issue with the system.
The AI failed to understand the commercial context of the campaign.
Someone searching for emergency plumbing services does not want to land on a recruitment page.
This type of mistake could waste advertising spend and damage conversion rates.
Attempting to Correct the AI
A prompt was entered asking Gemini to generate more relevant service-based site links.
The response?
Essentially nothing useful.
The AI instructed the advertiser to manually edit the ads instead of generating improved suggestions.
This exposed another major weakness:
The system presents itself as conversational AI, but it does not actually behave conversationally in any meaningful way.
Verdict on Site Links
Fail
The AI lacked contextual understanding of campaign intent.
Is Google’s Gemini AI Ready for Advertisers?
Based on this test, the answer is currently:
Not Yet
While some parts of the system show promise, the overall experience still feels unfinished and extremely limited.
The strongest areas were:
- Website scanning
- Basic headline generation
The weakest areas were:
- Keyword research
- Descriptions
- Site links
- Conversational functionality
- Campaign context awareness
Should Advertisers Be Worried About AI Replacing Agencies?
A lot of advertisers and agencies are concerned that AI tools like Gemini could eventually replace PPC specialists.
But this test suggests there is still a very long way to go.
At present, the AI struggles with:
- Strategic thinking
- Contextual judgement
- Persuasive copywriting
- Campaign structure
- Conversion-focused optimisation
In reality, experienced PPC professionals still provide enormous value that AI currently cannot replicate.
Final Verdict: A Promising Idea With Poor Execution
Google’s vision for AI-powered campaign creation makes sense.
In theory, having AI assist advertisers with setup, copywriting, and asset generation could massively improve efficiency.
However, in its current state, Gemini inside Google Ads feels more like an unfinished beta test than a revolutionary tool.
Overall Scores
| Feature | Verdict |
|---|---|
| Website Scanning | Pass |
| Keyword Generation | Fail |
| Headlines | Partial Pass |
| Descriptions | Fail |
| Images | Partial Pass |
| Site Links | Fail |
For now, advertisers should treat Gemini as a minor assistant rather than a replacement for proper campaign management.
Human strategy, keyword research, and persuasive ad writing are still essential for successful Google Ads performance.
As AI continues to evolve, this may eventually change — but right now, Google’s Gemini integration is far from ready to take over campaign creation entirely.
