Thought Leadership: 4 Formats and a 5-Step Process

Nobody needs another brainstorm. The publishable material is already sitting in your calls and tickets.

Here is the test we apply before anything goes out under a client's name: could a smart competitor read this and say "no, that's wrong"?

If they can't, it is a summary. Summaries have a job. Not this one.

This page covers the narrow thing: having a position and publishing it. The wider programme it sits inside (topics, funnel, channel mix, budget) is in our guide to B2B content marketing, and shaping a story once you have one belongs to brand storytelling. Neither page tells you what to have an opinion about. This one does.

What thought leadership actually is

A piece of thought leadership makes a claim about how your industry works, says why the common view is wrong or incomplete, and puts something on the table the reader could check. That is the whole definition. The rest is production detail.

What usually gets published instead:

• The recap: a summary of what three trade publications already covered, with no line added.
• The definition post: "What is X?" written by a company that sells X, which is a category page wearing a blazer.
• The trends list: twelve predictions, none falsifiable, none the author would bet money on.
• The sanitised case study: a real project with every number, constraint and mistake stripped out, so nothing is left to learn from. Customer-proof pieces have their own craft, covered in case study marketing.

None of that is dishonest. It is simply not an argument, and an argument is what the format is for.

Why most thought leadership fails

Buyers are reading. In the 2025 Edelman-LinkedIn B2B Thought Leadership Impact Report (Edelman and LinkedIn; 1,934 executives surveyed 17 March to 3 April 2025), 64% of the "hidden" decision-makers a vendor never meets said they spend over an hour a week on thought leadership, and 71% said it shows a vendor's value better than conventional marketing does.

The quality is the problem. In the 2024 edition of the same study (3,484 executives across seven markets including India, fielded 30 November to 14 December 2023), only 15% rated the quality of the thought leadership they consume as very good or excellent. Three failure modes account for most of that gap.

No position. The piece describes something everybody already agrees exists. "AI is changing search" is not a position. "Publishers blocking AI crawlers are defending a revenue line that stopped existing in 2023" is, because somebody will fight you about it.

No evidence. The claim is fine and the support is a statistic from a listicle that cites another listicle. If your only backing is somebody else's blog post, you are quoting, not leading.

Committee editing. A draft with a sharp claim goes to legal, sales, the product lead and the founder's chief of staff. Each removes the one sentence that made them uncomfortable. What publishes is approved and says nothing. The same 2025 report found 86% of hidden decision-makers want ideas that challenge their assumptions and 65% prefer a human, informal tone. Both are what four rounds of review sand off.

Where the raw material comes from

Founders say they have nothing to write about. They almost always have too much and no system for getting it out. The material sits in conversations that already happen in the business, and the extraction question matters more than the source.

Our observation: the last row produces the best piece and is the hardest to schedule, because it needs the founder unhurried. Book it as a call, never as "send me your thoughts".

The four formats that work, and when to use each

• The position piece (roughly 800–1,200 words). For when you have a view and no data. The claim goes in the first three lines, and the obvious objections get answered in the body rather than ignored.
• The original data study. For when you hold information nobody else can get. It earns the links and citations, and most companies skip it because it is work. State the sample size and method, and publish the rows so a reader can check one against reality.
• The worked example. For a specific outcome you can describe honestly, including the part that went wrong. Numbers withheld for confidentiality is fine. Numbers withheld because they were unimpressive is how you get the case study nobody finishes.
• The teardown. For when an industry default is wrong and you can show why. Attack the practice, not a named company, unless you are ready to be argued with in public by that company.

Four is enough. A format list of twelve means nobody has decided what they think. When we build a content marketing programme for a B2B client, one flagship data study a year and a monthly position piece carry most of the weight.

How to turn one sales call into one publishable piece

This is what we run when a founder says they have nothing to say. It starts from a call that already happened.

1. Pick the call you lost, not the win. Losses contain the sentence you couldn't answer, and that sentence is the article. Get the recording or transcript.
2. Find the line that stopped the room. One objection, one moment of hesitation, one question the rep bluffed. Mark it. That is your headline claim, inverted.
3. Write the answer you wish you had given, in full, as an argument rather than an FAQ entry. Length follows the idea. If it runs to 400 words, it was a tweet, not an article.
4. Add the evidence you actually hold. How often does the objection come up across your accounts? What happened in the two cases you may describe? If the honest answer is "twice, and once it went badly", write that. It beats a number you cannot source.
5. Get one hostile read, not a committee. One person whose job is to attack the argument and who is allowed to say "this is wrong". Fix what breaks, then publish before anyone else softens it.

Our working estimate for the expert's time across those five steps is about 90 minutes: a 30-minute recorded debrief, a 20-minute review and a 20-minute argument. That comes from how we schedule the work, not from a benchmark.

Founder voice and ghostwriting: where we draw the line

Our view, and how we handle it at Digital Hangover. Ghostwriting a founder's thought leadership is normal and, done properly, honest. The line is about ownership of the idea, not authorship of the sentences.

From the named person: the position, the specific claims, the anecdotes, and the willingness to defend them in the comments. Not from them: sentence construction, structure, headline, edit.

Three rules we hold. We will not publish a position the named author has not read, argued with and approved. We will not invent an anecdote or attribute an experience the person did not have. If a founder has no view on a topic, we change the topic rather than manufacture one, which occasionally costs us a publish date.

Google's framing is useful here. Its creating helpful, reliable, people-first content documentation (last updated 10 December 2025) asks who created the content, how and why, and whether it "clearly demonstrate[s] first-hand expertise and a depth of knowledge". A ghostwritten piece passes when the expertise is real and attributed, and fails when the byline is decoration.

A publishing cadence a busy executive can hold

One substantial piece a month, plus a short weekly post reacting to something real, is a rhythm most founders sustain for a year. Two long pieces a week is a rhythm most abandon by week six. Our observation from running founder-led programmes, not a study.

Cap the founder's input rather than the output. Their commitment is 30 minutes of recorded talking and 20 minutes of review per piece; everything between is somebody else's job. Book the recording as a recurring calendar meeting, because "write the post" never survives a busy quarter.

On distribution: LinkedIn is where B2B thought leadership gets read in India, and the organic mechanics are in our LinkedIn marketing guide. One honest note. We could not verify LinkedIn's current article and newsletter publishing steps this run, because LinkedIn's robots.txt disallows automated fetchers on its Help Centre pages. This page therefore makes no claim about how those features work today. Check them in LinkedIn's Help Centre before building a process around them.

How to measure thought leadership honestly

Thought leadership influences who gets shortlisted, not who converts this month. Measure whether the right people in your target accounts read it, whether sales uses it inside live deals, and whether inbound conversations quote your argument back. Pipeline attribution in month one is a promise nobody can keep.

• Reach into the buying group, not reach overall. Ten reads from four target accounts beat 4,000 impressions from people who will never buy.
• Sales-cited use. How often a rep sends a specific piece into an open opportunity. Ask monthly and log it; no analytics tool reports this.
• Inbound quality. Does the meeting request reference the claim you made? That is the argument doing the work, not the SEO.
• Brand mentions in AI answers. Whether assistants cite you when someone asks about your category. The mechanics are in our guide on ranking in AI Overviews.

The 2025 Edelman-LinkedIn report helps when you set expectations with a board: 79% of hidden decision-makers said they are more likely to champion an RFP response from a company producing quality thought leadership, and 95% said it makes them receptive to sales outreach. Those are stated attitudes from a survey, not attribution measured in your CRM. Quote them as evidence that the channel matters, never as a forecast of next quarter.

Why original data is the version that survives AI search

If you publish one thing a year, make it something only you could have published. Google's guide to optimising for its generative AI features (last updated 10 July 2026) contrasts a first-hand review that "provides a unique perspective based on personal experience" with "a summary of existing content [that] simply restates information already available elsewhere", and asks for "unique expert or experienced takes that go beyond common knowledge and the ordinary".

The same page is blunt about the shortcuts being sold: you do not need to "create new machine readable files, AI text files, markup, or Markdown" to appear in Google Search including its AI features, and llms.txt files neither help nor harm because Search ignores them.

A worked example from our own desk. In September 2026 we read the robots.txt file of 99 Indian companies in one day (the NIFTY 50, 30 D2C brands, 20 news publishers) to answer a question nobody had checked: are Indian brands blocking AI crawlers the way Western publishers did? 78 files were readable, and exactly one of the 71 corporate and D2C sites blocks an AI crawler. Of the 23 files naming an AI bot at all, 18 say Allow. The finding was not clever. It was just unavailable to anyone who had not read the files. That study publishes shortly on our research section, and this page will link to it then.

No markup substitutes for having something to say. You need one question you can answer from information you already hold, and the discipline to publish the answer even when it is duller than you hoped.