Across this category cut, the AI writing tools most cited by AI engines were copy.ai (5 citations), then jasper.ai, writesonic.com, and grammarly.com tied at 4 citations each. Those three names sit at the top of a short, vendor-heavy list. The numbers come from one slice of my AI-Search Visibility Benchmark: 6 queries, 53 citations, and 33 distinct domains, drawn from a wider run of 1,055 citations across 508 domains and 148 queries on Perplexity and ChatGPT. The sample is small, so treat the ranking as directional.
I’m Andrii Byzov, a fractional CMO for B2B tech. I run AI search optimization for B2B SaaS, and this is one category from the broader AI-Search Visibility Benchmark that I keep returning to, partly for the meta-note: these are AI engines deciding which AI writing tools to recommend.
Key takeaways
- Copy.ai led with 5 citations, ahead of jasper.ai, writesonic.com, and grammarly.com at 4 each.
- The category ran on 6 queries, 53 citations, 33 domains across Perplexity and ChatGPT.
- Vendor-owned domains dominate the top of the list, with review and reference sites layered in below.
- The honest read: 6 queries is a thin sample, so the order can shift, and any single count is directional.
The ranked top domains
Here is the top of the cited-domain list for AI writing tools in this cut. Counts are raw citations within the 53 the category produced.
| Rank | Domain | Citations |
|---|---|---|
| 1 | copy.ai | 5 |
| 2 | jasper.ai | 4 |
| 2 | writesonic.com | 4 |
| 2 | grammarly.com | 4 |
| 5 | claude.ai | 3 |
| 5 | rytr.me | 3 |
| 7 | aiunpacking.com | 2 |
| 7 | notion.so | 2 |
| 7 | techradar.com | 2 |
| 10 | seojuice.com | 1 |
The gaps here are small. One citation separates first place from the three-way tie, and a single extra run could reorder the middle of the table. That is the nature of a 6-query slice, and it is the first caveat I would hand any vendor.
The pattern, and what it suggests
The clearest signal is that vendor-owned domains lead. Copy.ai, jasper.ai, writesonic.com, grammarly.com, rytr.me, and claude.ai are all product sites, and together they take most of the top of the table. The review and reference layer (techradar.com, aiunpacking.com, notion.so, seojuice.com) sits below them and looks thinner here than in some other categories I have seen.
If that holds with more data, it points to a useful idea for an AI-writing vendor: in this category the engines were willing to pull from the tools’ own pages, not only from third-party roundups. Your product, feature, and comparison pages are not just sales assets in that world. They are candidate answer sources. That is the core of generative engine optimization: writing pages an engine can lift a clean, accurate answer from, then earning the authority that makes the engine trust them.
There is a meta-note worth naming, kept grounded in the data. This is AI deciding which AI writing tools to recommend, and in this slice it leaned on the vendors’ own descriptions to do it. That rewards vendors who keep their pages clear, current, and answer-shaped, and it means a stale product page is a missed citation, since the engine had fewer outside sources to fall back on here.
What an AI writing vendor should take from this
A few implications, held loosely given the sample size:
- Treat your own pages as cite-worthy. The engines pulled from vendor domains here, so the clarity and structure of your product and comparison pages matter for whether you get named.
- Do not ignore the review layer. Techradar, notion.so, and the smaller reference sites still appear. Presence in the roundups that cover your category adds surface area for a citation.
- Measure the category, not a single query. Six queries shows a shape, not an order. Track citations across more runs before acting on any one rank.
- Build the foundation. Entity clarity and structured data make a page easier to lift from. See what is AI search optimization for where that starts.
The honest caveat
This is one category from a larger benchmark, and 6 queries is a small base. The counts are close enough that a few extra runs could change the order, so read any single number as directional, not precise. The broader benchmark runs 148 queries and 1,055 citations, which is where I put more weight. No one controls how these engines choose sources, so the work improves your odds of being the cited answer over time, not your certainty of it. That is the realistic promise behind AI search optimization. If you want to talk through where your category stands, I’m on LinkedIn.
FAQ
Which sources do AI engines cite most for AI writing tools? In this slice, copy.ai led with 5 citations, then jasper.ai, writesonic.com, and grammarly.com tied at 4 each. The cut covered 6 queries, 53 citations, and 33 domains across Perplexity and ChatGPT. Read it as directional given the small sample.
How big was the AI writing tools sample? It drew 53 citations from 33 distinct domains across 6 queries, pulled from the wider benchmark of 1,055 citations, 508 domains, and 148 queries. Six queries is thin, so the rankings can move.
What pattern shows up in who gets cited? Vendor-owned domains lead, with copy.ai, jasper.ai, writesonic.com, grammarly.com, rytr.me, and claude.ai all product sites, plus a thinner layer of review and reference pages like techradar.com and notion.so. The engines leaned on vendors’ own pages here, though the sample is small.
What should an AI writing vendor do with this? Treat your product and comparison pages as cite-worthy answer sources, and earn presence in the review and reference sites that round out the list. The goal is improved odds of being the cited answer over time, not a guarantee.