TL;DR
- Expert material can help in AI search when it states facts, methods and limits that answers and buyers can use. These surveys do not establish an effect on AI citations.
- In a 2025 Edelman–LinkedIn survey, 64% of “hidden buyers” trusted thought leadership more than marketing materials.
- In Orbit Media’s 2026 survey, marketers who publish research more often said they got strong results. That is what they reported, not proof of cause.
- Publish expert proof, such as a testing memo with data and limits, rather than another generic positioning piece.
The short answer: proof helps, labels do not
Thought leadership can help, but only the kind that carries proof. A piece that restates a trend in confident language gives a buyer little to check. A piece that shows how your team tested something, what it found and where the finding stops gives the buyer specific material to use. It gives any system that summarizes the topic the same thing.
That second kind is plausibly useful in AI search. Answers need specific statements they can attribute, and many generic pieces lack them. This is a reasoned hypothesis, not a measured effect. Neither study reviewed here isolates an effect on AI citations or rankings. The two studies below measure what buyers say and what marketers report. For how answers select sources, see our citation diagnostic guide.
What buying groups say they value
The 2025 Edelman–LinkedIn report studies “hidden buyers”: final decision-makers, such as finance or procurement staff, without deep product knowledge. “Target buyers” are the product experts. The table further down gives the survey method.
64% of hidden buyers trusted thought leadership more than marketing materials (target buyers: 65%). 63% read it for over an hour a week (target buyers: 64%). The chart adds three related answers from hidden buyers.
Edelman–LinkedIn, U.S. LinkedIn member survey, March–April 2025. Same question, different buyer roles; subgroup sizes are not stated. This question does not measure AI search. 2025 Edelman–LinkedIn B2B Thought Leadership Impact Report (PDF).
What the survey means for a content plan
These are stated attitudes, not observed buying behavior. The sponsor sells B2B marketing, so treat the results as useful but not neutral. The survey says nothing directly about AI search.
It does show who reads expert material in a deal. A finance or procurement reviewer may never meet your sales team, but may read a well-argued memo. That reader needs evidence about cost, risk and limits, not a vision statement.
Does publishing original research pay off?
Orbit Media’s 2026 blogging survey, published September 8, 2026, asked 1,042 content marketers about their work and results. 37% said they publish original research, down from 49% the year before.
Among those research publishers, 22% reported “strong results.” Across all respondents, only 13.9% did.
These are the marketers’ own reports, so they show a link, not a cause. “Strong results” is each marketer’s own judgment, and the page does not explain how people were recruited. Teams that publish research may also have bigger budgets, better editors or stronger brands. The survey does not measure AI citations.
What each study can and cannot tell you
The two surveys come from different groups and ask different people different things. Check the scope of each before you rely on it.
| Point of comparison | Edelman–LinkedIn (2025) | Orbit Media (2026) |
|---|---|---|
| Who answered | 1,934 executives surveyed via LinkedIn; U.S. market; group sizes not published | 1,042 content marketers; recruitment not described |
| When | Fielded March 17 to April 3, 2025 | Published September 8, 2026; field dates not stated |
| What was measured | Buyers’ stated views | What marketers do and how they rate their own results |
| Relevant result | Trust it over marketing materials: 64% of hidden buyers, 65% of target buyers | Strong results: 22% of research publishers, 13.9% of all respondents |
| Who ran it | Edelman with LinkedIn | Orbit Media, an agency that publishes its own research |
| Cannot show | Whether thought leadership changed a purchase or an AI answer | Whether research caused the results, or any effect on AI citations |
The surveys cannot be merged into one effect. Together they back a narrower claim. Buyers say they value solid expert material. Marketers who publish research say they get strong results more often. Neither shows that thought leadership raises AI visibility.
Expert proof or a positioning piece?
Use this comparison to decide whether a planned piece deserves a place in the calendar. It is an editorial test, not a research finding, and the example claims are illustrative.
| Check | Generic positioning piece | Expert proof |
|---|---|---|
| Main claim | “The industry must rethink reliability” | “In our bench tests, wear rose sharply above a stated solids level” |
| Evidence | Opinion, trend talk and a logo | Method, conditions, results, limits and a named author |
| Who in the buying group can use it | Mostly the person already convinced | Engineers check fit; procurement compares bids; finance weighs maintenance cost |
| What an AI answer could use | Little that is specific to you | Specific statements with conditions, attributed to your team |
| Main risk | Sounds like everyone else | Overclaiming if limits are cut during editing |
Worked example — fictional: from test data to a procurement-ready memo
A fictional maker of mechanical seals for slurry pumps wants a thought-leadership piece. The first draft is called “The future of pump reliability.” It cites market trends and ends with a demo request. No buyer could use it to choose a seal.
The engineering team has something better: bench tests of seal-face wear in abrasive slurry. The editor turns those tests into a memo. The table shows what the memo must contain before it is published.
| Memo field | Completed illustrative entry |
|---|---|
| Buyer question | How long will this seal last in our slurry, and which conditions shorten its life? |
| Readers in the buying group | Maintenance engineer (fit), procurement (comparing bids on total life), finance (service intervals), safety lead (leak risk) |
| Data required | Test method, slurry type and solids range, pressure, temperature, speed, number of runs, failure rule, test dates and who ran the tests |
| What the memo states | Measured wear under each tested condition, shown in a table, with the method summarized in plain language |
| Limits stated beside the results | Bench results, not field life; untested slurries and speeds are out of scope; not a warranty |
| Use for procurement | A list of conditions any supplier should state when quoting seal life, so bids can be compared on equal terms |
| Approvals | Test engineer checks the data; quality manager checks the method; a named reviewer checks every claim against the test record |
| Format | A web page with tables and a named author, linked from the relevant product page |
| What was cut | A headline promising longer seal life with no comparable baseline |
The memo is less dramatic than the first draft, but every reader in the buying group can now use it. It also gives search and AI systems specific statements they can attribute. Whether any answer uses them is something to observe, not to promise.
How to judge whether expert material is working
Judge the memo first by business use. Do sales and application engineers send it to buyers? Do procurement questions arrive better prepared? Does the same objection stop coming up in calls?
Then observe AI answers without assuming a cause. Record a fixed set of buyer questions before and after publication. Label whether answers mention the company, cite the memo and state its findings accurately. Our AI visibility measurement guide explains the counting rules. Note the publication date next to other marketing activity, so a later change is not credited to the memo alone.
Thought leadership and AI search: common questions
Do we need original research to appear in AI answers?
The studies reviewed here do not show that it is required. Original data gives you something specific to say that others cannot, which helps buyers. Whether an AI answer uses it depends on the question, the platform and the competing sources.
Is a LinkedIn post by our CEO thought leadership?
It can be, if it shares an insight others can check. Short posts help spread an idea, but the full method, data and limits should live on a page you control. There, buyers and search systems can find the whole context.
How do we stop experts from overclaiming?
Write each limit next to its result, not in a footnote. Ask the person who ran the work to approve every number and condition. If marketing wants a stronger headline, the evidence has to support it first.
How long until AI answers reflect a new piece?
There is no reliable timeline. Crawling, indexing and answer generation differ by platform and change over time. Set a review date, rerun the same questions and record what changed without assuming the piece caused it.
Connect your experts’ material with buyer questions
Many B2B companies already hold strong expert material, such as test records, field notes, method notes and engineers’ answers to buyer questions. SemShift’s GEO work starts from the questions buyers ask. We connect them to material your experts have approved and improve how your pages present that evidence. Your team owns the data and approves the claims; we help make it easier to find and understand.