Which marketing tools the models recommend
We ask the answer engines the questions buyers ask, and publish who gets named. The prompts and the raw source lists are here so you can run it yourself.
Nine buyer questions, asked in Perplexity because it prints its sources. The pattern across all nine: head terms are held entirely by high-authority publishers, and the specific questions are answered by whoever happened to write about them, including personal blogs. On "Arcads alternatives" all ten cited sources were rival tools ranking themselves. On what tools cost, every cited source was an unknown blog. Nobody has authority in this category yet, which is the finding.
How is this measured?
By hand, in Perplexity, because it prints its source list next to the answer. ChatGPT and Google AI draw on a broadly overlapping pool but do not show their working, so Perplexity is the readable proxy rather than the whole picture.
Nine prompts, each phrased the way a buyer would type it rather than the way a marketer would title a post. Sources recorded verbatim, no attempt to influence the result, re-run quarterly because the sets move. This is a snapshot of who is being read, not a quality ranking of the tools. A tool absent from a citation set may be better than every tool in it.
Which questions are already locked?
The head terms. "Best AI marketing tools for a solo founder" and "guide to the modern marketing landscape" return almost nothing but high-authority publishers: Zapier, HubSpot, Hootsuite, Improvado, Gartner, Deloitte, Kantar, the AMA.
One independent broke through across the whole set: Marketer Milk, cited on two of the nine. That is the useful signal here. A single-operator site can rank against Gartner on these questions, but not by writing a better list of ten tools. It happens by being the source other cited pages point at.
Which questions is nobody answering well?
The specific ones, and the gap is wide. Asked what tools cost per month, every cited source was a blog nobody has heard of. Asked how to get from zero to a thousand a month, the cited sources included a Blogspot post and two personal sites.
"Arcads alternatives" · 10 of 10 citations were rival tools
Every source was a competing product ranking itself first. The softest ground we measured, and the reason a neutral comparison page with honest pricing is worth writing.
"How much do AI marketing tools cost" · all unknown blogs
No recognisable source in the set. A price table with real numbers and a visible last-checked date is the obvious thing nobody has published.
"0 to $1k MRR" · a Blogspot post and personal sites
The weakest field in the set. Plenty of content exists, none of it has authority, so citations go to whoever is structurally easiest to quote.
"Cheapest AEO tool" · the winner was a price comparison
A review of thirty tools with real numbers took the citation. Consistent with the rest: facts get cited, opinions do not.
"Directory of AI marketing tools by what they do" · we were absent
Cited: Marketingtoolz, ColdIQ, TheCMO, ai-for-marketing, Improvado. That prompt is this site’s entire positioning and we were not in it.
What shows up in almost every answer?
Reddit and YouTube. They appear across most of the nine sets, and neither is a page anyone owns. On the UGC and clipping prompt, YouTube was cited five times and Reddit twice, against no directories at all.
The practical read is that a genuine answer in the thread where a question already gets asked, and one plain walkthrough video, will show up in citation sets faster than a new article. That is uncomfortable if you would rather write pages, and it is what the data says.
Good to know
Which AI marketing tools do the models recommend most?
For head terms, the models mostly name the publishers rather than the tools: Zapier, HubSpot, Hootsuite and Improvado dominate. For specific jobs the named tools change with whoever wrote the most quotable page, which is why the same question asked two ways returns different tools.
Does being in an AI answer mean a tool is good?
No. It means the page describing it was easy to quote. On one prompt every single cited source was a competitor ranking itself first, which is a fact about publishing incentives rather than product quality.
Can I run this myself?
Yes, and it takes about twenty minutes for ten prompts. Ask them in a tool that prints its sources, record whether you appear, repeat monthly. Free command-line options exist if you bring your own API keys.
How often is this updated?
Quarterly. The citation sets move enough that anything older than a few months is a historical record rather than a current one, which is why the date is on the page.
This expands the Summit camp on the climb, which lays out the stages either side of it. Tools by job are on the route index, the free ones are here, and the other answers are in insights. Updated 30 July 2026.