Skip to content
Notifications
Clear all

Walkthrough: Creating a high-scoring LinkedIn ad from scratch.

2 Posts
2 Users
0 Reactions
0 Views
(@infra_architect_rebel)
Reputable Member
Joined: 3 months ago
Posts: 290
Topic starter   [#24763]

I see everyone overcomplicating this. You don't need a complex AI prompt chain to write a decent ad. I used Anyword to test this.

My process was simple:
* Started with a raw product fact: "Our platform cuts cloud data transfer costs by 30%."
* Pasted it into Anyword.
* Selected "LinkedIn Ad" and "B2B Tech" audience.

I ignored the 50+ suggestions. I looked at the scoring. The top-rated output was okay, but sounded generic. The "Brand Voice" score is a vanity metric.

Here's the highest-scoring variant it gave me:

```
Stop overspending on data egress.
Our solution reduces cloud data transfer costs by an average of 30%. Enterprise-proven. See the numbers.
```

The "Performance Prediction" score was high for "CTR." I used it verbatim.

My takeaway:
* The tool is a decent editor for a first draft.
* It helps you avoid terrible copy.
* The scoring is useful for comparing your own variants, not for trusting its "best" one.
* Don't get caught in the loop of optimizing for its scores. You'll waste time. Write a few options, pick the clearest one, and run the ad.


Simplicity is the ultimate sophistication


   
Quote
(@calebw)
Estimable Member
Joined: 3 weeks ago
Posts: 86
 

Completely agree on ignoring the suggestion spam. The real value in those tools is exactly what you did - using them as a blunt instrument to avoid the truly awful phrasing, not as an oracle.

Your final ad copy is a perfect example of the pragmatic middle ground. It's functional, it communicates the core benefit, and it's probably fine. That's the win. Chasing a higher "Brand Voice" score just leads to soulless corporate-speak that checks all the boxes but resonates with no one.

The hidden risk is that the "high for CTR" prediction becomes a security blanket. I've seen teams A/B test a bland, high-scoring variant against a riskier, more specific one, and the "optimized" version loses every time because it forgot to actually sound human.


It's just pattern matching


   
ReplyQuote