GeneralMind is a Berlin company, founded in 2025, that automates document-heavy order and procurement work with AI agents. It handles purchase-order confirmations and follow-ups, invoice and claims matching, and sales-order intake. It reads whatever suppliers and customers send (email, PDF, Excel, fax-to-email, scans) and writes the result into ERPs such as SAP, Oracle and Microsoft Dynamics, working alongside the ERP rather than replacing it.

Is this a fair comparison?

Both sets of pages sit on the same domain, carry August 2026 publication dates, and are measured by the same daily pipeline against the same 197 questions. If anything, the HubSpot set had the advantage: six of its seven articles went live on 11 August, six days before the guide, so they had longer to be cited. What differs is how the guide was aimed: at a gap the tracking had measured, written in the structure, style and method AI actually cites.

Which pages count as HubSpot content?

Every GeneralMind blog page dated August 2026 other than the guide: seven articles, each in German and English, fourteen pages in all. Dates come from each page’s own published-date markup.

How do you know the guide caused the topic to move?

Ten of the 197 monitored questions sat under the treated topic, and the other 187 were left untouched as a control over identical windows. The treated questions gained 11.3 percentage points, the untreated ones 0.5. The control is slightly generous, because the guide was also cited on some untreated questions.

How was this measured?

All figures come from GeneralMind’s Asky data: 197 monitored questions across 12 topics and 53 sub-topics, in German and English, for the German and US markets. Each question runs daily against ChatGPT, Google AI Overviews and Perplexity, with 23 tracked competitors. None of the 197 questions names GeneralMind. The treated cohort is the ten questions under the procure-to-pay topic, tracked from 5 August. Before is 5 - 17 August (11 measured days, 330 answers). After is 21 August - 4 September (13 measured days, 390 answers). The final week is 29 August - 4 September (5 measured days, 150 answers). The per-page comparison counts citations to each URL across all 197 questions from 31 July to 4 September. Each language edition counts as its own page, for the guide and the HubSpot set alike. Citation rate is the share of captured AI answers linking to at least one GeneralMind page; see how AI citation tracking works. Every per-domain figure is measured on the ten treated questions and is not a site-wide or category-wide ranking. Third-party sources other than HubSpot are described by category rather than named.

Case study: one GeneralMind guide earned 9x the AI citations

One Asky guide vs a month of HubSpot posts: 9x the citations

GeneralMind, which automates order and procurement work with AI agents, published blog content two ways in August 2026. Seven articles, each in German and English, were produced in HubSpot; one guide was written in Asky, aimed at a topic the tracking showed nobody owned. Measured against the same 197 buyer questions, none of which name GeneralMind, the Asky guide drew 18.5 citations per page and the HubSpot pages 2.0. On the ten questions the guide targeted, the share of AI answers citing GeneralMind rose from 6.7% to 17.9%.

How did one guide compare with a month of HubSpot content?

The guide earned more citations on two pages than the other fourteen August pages earned together. A citation here is one AI answer linking to one page. The guide went live on 17 August and was first cited in an AI answer on 21 August, four days later. By 4 September its German and English editions had drawn 37 citations across two pages. The fourteen other blog pages GeneralMind dated August drew 28 across fourteen, and nine of the fourteen were never cited at all.

In the final measured week the guide drew 19 citations. The other fourteen pages drew 8.

Blog content dated August 2026, citations 31 Jul - 4 Sep	Pages	Citations	Per page	Final week (29 Aug - 4 Sep)
Written in Asky: one guide, German and English	2	37	18.5	19
Produced in HubSpot: seven articles, German and English	14	28	2.0	8

The difference is not volume. It is aim. The guide was written against a topic the tracking had already shown to carry demand with no clear owner.

Before publication, even the most-cited sources appeared in only about one AI answer in four on those questions. Its German edition drew all 19 of its citations from the German-language questions in that set. This is the approach described in using AI citation gaps to steer content strategy.

“In the same month my colleague had prepared about a dozen blog posts in HubSpot. The one article we wrote with Asky, aimed at a topic that came out of the data, gets cited roughly nine times as often per page. The difference isn’t volume, it’s knowing where it was worth writing and what to write.”

Tom, GTM Engineer, GeneralMind

Did the targeted questions move more than the rest?

Yes, by a wide margin. Ten of the 197 monitored questions sit under the procure-to-pay topic. Citation rate is the share of AI answers that link to at least one GeneralMind page. On those ten questions it went from 6.7% before publication to 17.9% after, and 21.3% in the final measured week.

The other 187 questions are the control. No new content was aimed at them, and their citation rate rose from 0.9% to 1.4%. With part of that rise coming from the guide itself.

Cohort	Questions	Before (5 - 17 Aug)	After (21 Aug - 4 Sep)	Change
Treated, the procure-to-pay topic	10	6.7%	17.9%	+11.3 pp
Untreated, everything else in the tree	187	0.9%	1.4%	+0.5 pp

So about 10.8 points are attributable to the content, and after publication the treated questions were cited at almost 13 times the rate of the untreated ones.

The rise came from all three tracked AI engines, not one. Engines differ in how they pick sources, so a gain on all three is the stronger result.

Engine, treated questions	Before (5 - 17 Aug)	After (21 Aug - 4 Sep)	Change
ChatGPT	0.9%	9.2%	+922%
Google AI Overviews	3.6%	10.8%	+200%
Perplexity	15.5%	33.8%	+118%

Where does GeneralMind rank among cited sources now?

Sixth of 352 on the targeted questions, up from 26th. Before publication, AI answers to those ten questions drew on 434 different domains, and GeneralMind was the 26th most cited, with 23 citations. After publication it was 6th of 352, with 80. In the final measured week it was 4th, and the guide was its most-cited page for those questions.

Worth noting what it is up against. The source at the top of that list is an enterprise resource-planning vendor’s own documentation, cited 308 times across 118 pages, or 2.6 citations per page. Thirteen GeneralMind pages at 6.2 apiece is a different kind of position: engines returning to specific pages as the answer rather than sampling a large site.

“Generating an article is easy now. Knowing what’s worth writing, and creating something AI engines actually cite, is the hard part. That’s where Asky stood out: the article we wrote with it became our most-cited blog content.”