Back HexScope Lens

All Six AI Assistants Reach For LinkedIn On Work Questions

POINT Key points
  • LinkedIn outranks every company's own site on work questions
  • What gets quoted moved from profiles toward posts and long articles

“Social Is Awareness Budget. AI Is A Separate Line.”

Ever watched a planning meeting tidy the year into those two boxes, with nobody in the room pushing back?

Structured data on your own site goes in the AI box. The social calendar goes in the awareness box. It’s clean, it survives contact with finance, and I built that exact table more than once — one half of the spend aimed at machines, the other half aimed at people.

Then somebody went and counted where the assistants are actually pulling their sentences from when the question is about work, and the line between those two boxes stopped holding.

The most-quoted source wasn’t a company’s own site. It was a social platform.

Three months, 1.4 million citations, six assistants.

Which Domain Do The Models Quote On Work Questions?

LinkedIn — and on every single one of the six platforms measured, which is the part that makes it hard to wave away as one model’s quirk.

The six were ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Microsoft Copilot and Perplexity. On ChatGPT’s overall domain ranking, LinkedIn climbed from about 11th to 5th over the same window (Profound, published 9 March 2026, covering 15 November 2025 to 15 February 2026).

Profound calls that the biggest shift in domain authority it observed all year. A slot that moves that far in twelve months is a slot worth watching.

A “citation” here just means a source the assistant leaned on while assembling its answer. Which is why this matters to you: a surface that isn’t yours at all can still be the thing feeding what the model says about your category.

The Part Of LinkedIn That Gets Quoted Has Moved

Inside that same three months, the team also split the LinkedIn citations by what kind of page got pulled:

  • Company and personal profiles: 33.9% → 14.5%
  • Posts: 20.9% → 26.0%
  • Long-form articles: 6.0% → 8.9%

Profiles dropped 19.4 points. Posts and long-form articles, taken together, went from 26.9% to 34.9% — up 8.0 points.

So the models have drifted off the business-card surface and onto the surface where somebody actually wrote something down. Since a profile is mostly a claim about yourself and a post is a claim you had to argue for, that drift is doing something a marketer can act on.

I’d been handing out beautiful business cards and never checking whether anyone turned them over. Uncomfortable reading, that.

Four Things To Say Out Loud Before This Goes In A Deck

Profound sells AI-visibility measurement. “Measure your AI citations and act on them” happens to be the conclusion that sells the product, so read the direction of the finding with that gradient in mind.

The two headline numbers also come from two different instruments. The domain ranking is built from real user conversations; the professional-query result is a structured measurement across a defined set of topics. They complement each other, and they aren’t the same metric, so don’t put them on one chart.

The measurement is English-language. Nobody counted what the composition looks like in Japanese, German or Spanish, and I wouldn’t assume the ranking transfers.

And it’s the vendor’s own panel, with no published third-party replication.

Drop those four and all that walks around the office is “apparently LinkedIn is number one”, which is how a real finding turns into a bad decision. I say them first, then the number.

Count How Much Of Your Own Subject Matter Is Readable Off-Site

Name the three things your company can genuinely speak to. Then go looking for writing on each of them that lives somewhere other than your own domain, and count what you find — posts from named people, proper long-form pieces, not the boilerplate paragraph on the company page.

If the count is zero, that’s the gap this data is pointing at. Nothing here says stop maintaining your own site; one more surface just turned out to matter.

There’s a second thing, and it’s the reason I’m writing this at all. Impressions on your own posts are visible to you, but which post an assistant leaned on while answering somebody’s question is not, because that happens on the model’s side of the glass. If you want to know whether any of this worked, that’s where the counting has to happen.


Sources

  • Profound, “LinkedIn is the most-cited domain for professional queries in AI search”, published 9 March 2026, tryprofound.com (1.4 million citations collected 15 November 2025 – 15 February 2026 across ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Microsoft Copilot and Perplexity. LinkedIn ranked first for professional queries on all six platforms and rose from about 11th to 5th in ChatGPT’s overall domain ranking. Within LinkedIn, cited profiles fell 33.9% → 14.5% while posts rose 20.9% → 26.0% and long-form articles 6.0% → 8.9%. Domain ranking and professional-query ranking are measured by two different methods and are not a single metric; the measurement is English-language only; the data comes from Profound’s own panel with no published third-party replication; Profound sells AI-visibility measurement.)
Share this article
Bluesky X
Back to all articles