4/
The clearest instruction was:
“Use Reddit and community discussions.”
That worked across all 5 queries.
When I asked Perplexity to use community evidence, it usually did.
That was the strongest source-steering signal in the test.
@firefliesai Funny you ask — I actually got distracted before choosing 😅
I started comparing AI meeting note tools, then turned it into a small test of how Perplexity decides which sources to cite in this category.
Happy to send the 75-search writeup + raw data if useful.
13/
I wrote up the full test here, with the data linked inside:
https://t.co/TVkQ6eanRl
Curious if others are seeing the same thing in their categories, especially where Reddit, YouTube, or comparison pages dominate the answer.
#GEO#AEO#AIsearch
12/
For brands, the practical question is not:
“How do we force Perplexity to cite us?”
It’s:
“What source layer does Perplexity already trust in our category, and do we exist there?”
Official pages matter.
But they are only one layer.
3/
Why repeat each search?
Because AI search is noisy.
If sources change once, you don’t know if your prompt caused it or if Perplexity was just doing normal Perplexity things.
So I repeated each condition to separate prompt effect from source churn.
1/
Can you control what sources Perplexity uses?
I ran 75 searches to test this.
Same category. Same queries. Different source instructions:
“use official sources”
“use reviews”
“use Reddit”
“avoid listicles”
The answer: you can nudge it. You can’t fully control it.
2/
I tested this in a real SaaS category: AI meeting note tools.
The queries were things like:
- best AI meeting note taker for sales calls
- Fathom AI review
- Fathom vs Fireflies
- Fireflies AI review
- Fireflies vs Fathom
Each condition was run 3 times in new searches.