Start here / 07 AI and search content
AI and search content · where the answers come from
One channel never carries visibility. In the tracked reading of EKU's buyer questions, one cited page in four hundred was on a social network.
When an assistant answers your buyer, whose pages is it reading?
Tracked reading of 1 to 2 October 2026 · studies read at their publishers on 2 October 2026
| Kind of source | Pages cited | What it is |
|---|---|---|
| Other company and trade sites | 164 | Makers, integrators and trade pages outside the tracked rivals. |
| Service provider sites | 110 | Dealers, integrators, rental and engineering firms. |
| Rivals' own sites | 74 | Product and application pages of the tracked rivals. |
| Knowledge bases | 45 | Technical documentation, standards bodies, government and reference pages. |
| Press | 3 | News and trade press articles. |
| Online stores | 3 | Parts and retail listings. |
| Social networks | 1 | Posts and profiles. |
| EKU's own site | 0 | No page of ekupd.com was cited. |
| Domain | Questions, of 37, in which it was cited |
|---|---|
| cat.com | 10 |
| comap-control.com | 10 |
| deif.com | 7 |
| cummins.com | 6 |
| vertiv.com | 5 |
| eaton.com | 5 |
| woodward.com | 5 |
| controlcomtech.com | 5 |
| siemens.com | 4 |
| unum-genset.com | 4 |
| se.com | 4 |
| generac.com | 4 |
Same reading. The one social page was on LinkedIn; the three press pages were two news sites and one technology site. No page of ekupd.com appears in the 400.
No single channel carries visibility. Every study below comes from a firm with something to sell, and they define their terms differently, so the figures are never averaged.
| Who published it, and when | What it reports | Read it with this in mind |
|---|---|---|
| Meltwater, 12 May 2026 | 9.5 million AI citations across 16 B2B categories and six AI models: LinkedIn holds 0.53% of citations, the second most cited single site after YouTube at 1.52%. | Research by a media-monitoring vendor, about LinkedIn. |
| Muck Rack, 7 May 2026 | Across more than 25 million cited links in 17 industries, earned media accounts for 84% of AI citations; journalism alone is 27%. | "Earned" includes journalism, academic, government, reference and third-party corporate pages. Published by a press-relations software firm. |
| Profound, 5 June 2025 | 680 million citations, August 2024 to June 2025: LinkedIn is 1.3% of one engine's citations and 0.8% of another's, and absent from a third's top ten. | One vendor's index; engines differ a lot from each other. |
| OtterlyAI, 8 September 2026 | 6.1 million citations, July and August 2026, United States, seven engines: in Industrial and Manufacturing, 77.4% of citations go to company websites. | The category is broad; not power controls on its own. |
| Zen Media, 30 September 2026 | 1,000 buyer prompts and 4,000 answers across the top 100 industrial-automation vendors: manufacturer websites show a 92% citation frequency; the average vendor's visibility is 4%. | A press release; the benchmark page's figures were not visible when it was read on 2 October 2026. |
| BrightEdge, 9 October 2025 | 98% of LinkedIn content gets no AI visibility; the 2% that is cited follows strict patterns, mostly long articles and courses. | No sample size printed. |
Label for every row: Observed at the publisher's own page, read 2 October 2026. None of these studies looks at power generation controls on its own.
A starting split for an industrial controls company. Shares of effort, by judgement.
In industrial questions, assistants lean on makers' own pages. This is where the buyer's words have to live.
Third-party pages carry most of what assistants cite. A case with figures is the entry ticket.
Where an assistant checks that a company exists and what it does.
Assistants quote video for how-to and monitoring questions.
Long, educational articles by named engineers are what gets cited. Short posts mostly are not.
Slow to earn. Worth one correct entry, kept up to date.
Working Hypothesis by Kanbina Lumina, from the strength of the evidence above. It is a judgement, not a measurement. Review the split at each monthly review, using dated publication and referral evidence; the tracked reading stays separate from the manual series.
Closing
An assistant answers from the pages it can find and trust. For EKU's buyer questions, none of those pages is EKU's yet.
Posting on one network reaches people who already follow EKU. The buyer who only knows the problem is reached through EKU's own pages and through third parties.
The marketing lead uses the split to plan each month's pieces. The tracked reading was taken with Kanbina Lumina's tool and is not repeated; EKU's own repeat is the manual wave on three assistants, in the lab card.
Validation status. Own reading: one AI assistant, 37 locked questions, 1 to 2 October 2026, read on the tracking screen by Kanbina Lumina on 2 October, Observed; not repeated. Outside studies: Observed at each publisher's own page on 2 October 2026; every publisher sells something. The split of effort is a Working Hypothesis.
Working foundation for executive review. Not approved external copy. Prepared for EKU Power Drives, 30 September 2026, by KanbinaLUMINA