Corrected 19 August 2026. The first version of this post counted product name variants as separate products, which inflated three rows and produced the wrong conclusion. The numbers below are recomputed from the same committed scan. What changed and why is at the end.
We ran five buyer questions about AI SEO tools through ChatGPT, Claude, Perplexity and Gemini: twenty answers, one day in July 2026, every named product counted.
The interesting thing is not who won. It is that the engines disagree in a pattern.
The category splits by engine, not by product
| Product | Mentions (of 20) | ChatGPT | Claude | Perplexity | Gemini | Best rank |
|---|---|---|---|---|---|---|
| Surfer SEO | 6 | 1 | 1 | 4 | 0 | #2 |
| SEOmatic | 4 | 1 | 1 | 1 | 1 | #1 |
| ZRank | 3 | 1 | 1 | 0 | 1 | #1 |
| Serpa | 3 | 1 | 1 | 0 | 1 | #2 |
| Contentpen | 3 | 1 | 1 | 0 | 1 | #1 |
| Outpacer | 3 | 1 | 1 | 0 | 1 | #3 |
| SEObolt | 3 | 1 | 1 | 0 | 1 | #1 |
| Byword | 3 | 1 | 1 | 0 | 1 | #1 |
| GrandRanker | 3 | 1 | 1 | 0 | 1 | #1 |
| Allable | 3 | 1 | 1 | 0 | 1 | #2 |
| Semrush | 3 | 0 | 0 | 3 | 0 | #1 |
| Ahrefs | 2 | 0 | 0 | 2 | 0 | #2 |
| Frase | 2 | 0 | 0 | 2 | 0 | #4 |
Read the Perplexity column against the other three.
Eight tools sit on exactly the same line: named by ChatGPT, Claude and Gemini, and never once by Perplexity. Three names sit on the mirror image: Semrush, Ahrefs and Frase, named only by Perplexity and by nobody else.
One product, SEOmatic, appears on all four.
Why the split is not random
Perplexity's Sonar searches the live web on every question. The other three answered these prompts with web search enabled too, but they lean far more on what the model already knows.
So the two groups are answering different questions. Perplexity is closer to "what does the web say about SEO tools right now", which surfaces the companies with the most existing coverage. ChatGPT, Claude and Gemini are closer to "what do I associate with this category", which surfaces whatever has been written in a form they absorbed.
That is why Ahrefs, a tool almost every SEO has used, is invisible on three of the four assistants your buyers actually open.
The tools winning the three-engine group have something in common
Look at what the engines cited while answering.
| Cited source | Citations |
|---|---|
| seomatic.ai | 4 |
| zplatform.ai | 3 |
| tryserpa.com | 3 |
| contentpen.ai | 3 |
| outpacer.ai | 3 |
| seobolt.io | 3 |
| surferseo.com | 3 |
| byword.ai | 3 |
| grandranker.com | 3 |
| allable.ai | 3 |
Almost every domain in that list belongs to a product in the three-engine group, and each is cited from its own site. These are not household names being remembered. They are companies whose own pages the engines read during the answer.
Several of them say plainly what they are doing. GrandRanker's homepage advertises content "optimized for AI search (Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini)". It is working, at least on three engines out of four.
SEOmatic is the only product that shows up everywhere, and seomatic.ai is the single most cited domain in the scan. That is the combination worth copying: cited enough for the search-grounded engine, and present enough in written form for the ones that are not.
What to do with this
If you sell an SEO tool, the useful question is not "am I visible in AI answers". It is "which half am I visible in".
If you only appear on Perplexity, you have coverage and no presence in the models. If you only appear on the other three, you have presence and no live coverage, and a model update can take it away. Both halves are fixable, and they are fixed by different work.
What changed in this correction
The original table read grandranker.com 8 mentions from 1 engine, byword.ai 7 from 1, allable.ai 5 from 1, and concluded the category was full of single-engine favourites.
Two defects produced that. Our extraction sometimes recorded a product under its display name and sometimes under its domain, so GrandRanker and grandranker.com were counted as two different products. And one Gemini answer listed the same three domains repeatedly down to position 20, which we counted as twenty recommendations rather than the parsing artefact it was.
Merging the name variants and dropping that one answer gives the table above. The three inflated rows are each 3 mentions across 3 engines, and the real pattern, which was there all along, is the Perplexity split.
We publish the underlying scan data with every post so that this kind of thing is checkable. This one was caught by an audit that now fails our build if any two product names in a scan collapse to the same company.
Method, honestly
Five buyer-intent prompts (for example "best AI SEO tool for a small SaaS" and "is an automated SEO agent worth paying for"), each asked once to ChatGPT (gpt-4o-mini), Claude (Haiku 4.5), Perplexity (Sonar) and Gemini (2.5 Flash) with web search on, same day. Mentions count answers, so a product named twice in one answer counts once. Names are as the engines stated them, merged across obvious variants. One-day snapshot, and rankings move.
Whoaisays tracks whether AI assistants recommend your brand, every week. Free for 14 days, no card required; plans from $29/mo.