A lot of blog drafts end up saying what every other post on the topic already says. The writer read the top few results, rewrote them, and never asked the person who actually does the work for a single real example. This guide is for content writers and the marketing leads who review their drafts, and it walks through eight steps that borrow journalism skills for content marketers can use right away: how to interview a source for a blog post, how to check what the source tells you, and how to show readers where each claim came from. By the end you'll have a process you can repeat on every post, and drafts that read as researched instead of recycled.
One thing to say up front. This guide covers interviewing and reporting only, so these are reporting techniques for content writing and not general storytelling advice. It doesn't cover persuasive writing, and it doesn't promise that a reported post will rank or get cited by AI engines. What it does give you is a way to add firsthand information and evidence to a post, which is the part most drafts are missing.
Why reporting makes a post feel researched
Reporting means getting information directly from people, documents, records, or your own observation, and then testing it before you publish. An interview is one way to do that, but an interview alone doesn't make a claim true. The improvement comes from asking better questions, checking the answers against evidence, keeping their meaning intact, and letting the reader see how you know.
A few terms will come up in the steps below, so it helps to have them straight now.
- A primary source is a person with direct knowledge, an original record or dataset, or original audio, video, or images. A secondary source is something that summarizes or interprets that material, like a news article or a book.
- On the record means what the person says can be published with their name attached. Off the record means it can't be published. On background means it can be used only under conditions you agree on ahead of time, often with a description of the person instead of a name.
- A direct quotation is the speaker's actual words. A paraphrase is the meaning in your words, and it still needs attribution.
Google's guidance on helpful content asks whether a page offers original information, reporting, research, or analysis and whether it adds real value beyond rewriting other sources. That is a quality test for your page. It isn't proof that any one interview will improve your rankings, and none of the sources behind this guide give a number for how much interviews help. Journalism tactics for SEO content work best when you treat the interview as a way to make the post better for the reader, and measure search and AI visibility separately afterward.
Step 1: Define the reader question and the claims that need reporting
Start by writing one question the post has to answer. Then list every claim a knowledgeable reader might push back on, and sort them into what your team knows and what it only assumes.
A simple reporting grid keeps this organized. Give it one row per claim and these columns: the claim, the person who might know, the original record, the verification status, and how you plan to attribute it. This is a working template, not an industry standard, so change it to fit your team.
Say you're writing about a content approval process. An employee explaining how the workflow runs is one kind of claim. A statement that the workflow cut approval time is a different kind, and it needs a measurement behind it. Putting them on separate rows keeps you from mixing them up later.
You're done with this step when every important claim has either a person who could speak to it firsthand or a document you can look at, and the ones you can't support yet are clearly marked as open.
Common mistake: starting from a keyword and asking an expert for general tips. That gets you another summary of advice that's already everywhere. Start from a question and the claims that would need proof.
If you track buyer questions in AI search, DeepSmith AI Visibility can show you the questions you're watching, the answers AI engines give, and the pages your competitors get cited for. That can help you pick a question that's worth reporting on. It doesn't tell you whether anything an interviewee says is true, and it doesn't run the interview for you.
Step 2: Choose people with direct knowledge and check their incentives
For each claim in your grid, find the person closest to the action. For a workflow that's the person who runs it, for a decision it's whoever made it, and for the experience of using a product it's someone who has really used it. If a claim needs specialist interpretation, add an expert on top of that, and look for original documents too.
Then check each person a little. What is their role, what have they worked on, do they have a commercial interest, and did they see the thing happen or hear about it from someone else? Poynter's guidance on judging a source suggests tracing information back to primary sources and thinking about expertise, funding, and bias.
You're done when you can say why each person is qualified to speak to each claim, what they can't know firsthand, and what relationship the reader should be told about. An employee is a good source on their own work and a self-interested source on how well the company performs, and both of those are true at once.
Common mistake: treating a big title as proof of firsthand knowledge, or treating two people on the same team as independent confirmation. If both are repeating the same company talking point, you have one source.
Step 3: Research the subject and prepare a flexible question route
Before the call, read the person's relevant work and any original material that's already public. Confirm names, job titles, dates, product terms, and how any metric is defined. Then build a prioritized list of topics, not a script. A script tends to make you stop listening.
A route that works well for most interviews looks like this:
- Orientation: "Walk me through what happened, starting with the first decision."
- Process: "Who did what next, and what did you use to decide?"
- A specific example: "Can you take me through one instance rather than the typical case?"
- Evidence: "Where could I verify that number or timeline?"
- Limits: "When did this approach fail, or when would you not recommend it?"
- Reader consequence: "What would a team trying this for the first time misunderstand?"
Put the most important factual questions first in your notes, and leave a little room beside each one to write down what you'll need to check afterward. Test your recorder and decide how you'll note timestamps. The Journalist's Resource advises setting an interview goal, prioritizing your questions, and reading the expert's work first, which is what this step is.
You're done when you have a short topic list, the claims you want to test, a first question for each topic, and a few follow-ups you might use. The source shouldn't be expected to write the article for you.
Common mistake: showing up only to get a quote. If that's the goal, you'll end up with an endorsement and nothing you can build a section on.
Step 4: Establish publication and recording rules before the interview
Start the conversation by saying who you are, what the piece is about, and how the person would like to be identified. Say plainly whether the conversation is on the record. If someone wants to stay anonymous or limit what you use, agree ahead of time on exactly what you can publish, whether you can quote them directly, and how you may describe them.
Ask permission before you record. Recording rules differ between jurisdictions, and calls that cross state or national lines can be trickier, so check current law and your company's policy. The Reporters Committee for Freedom of the Press has a reporter's recording guide that explains how consent requirements vary, though it isn't legal advice. If you can't record, take careful notes as you go, and never present a rough reconstruction as a word-for-word quote.
You're done when both sides understand the terms, you've confirmed the spelling of the person's name, their title, and their organization, and you've written down any restriction you agreed to.
Common mistake: assuming "off the record" means the same thing to everyone, then trying to sort it out after a revealing answer. Agree on the meaning first.
A note on anonymous sources. For an ordinary marketing how-to they should be rare. If an important claim can't be attributed, try to get it on the record or confirm it somewhere else. Don't quietly turn it into a fact the company asserts.
Step 5: Interview for the example beneath the generalization
Open with a broad invitation and then listen. Ask one question at a time. When an answer stays abstract, ask for a particular instance. Some questions that work well are "What was the last time that happened?", "What changed between those two steps?", "Who made that decision?", and "What record would show that?"
Another useful habit is to repeat a key word back as an open question. If someone calls a handoff "difficult," you can say, "You called the handoff difficult. What made it difficult?" When an answer is long or complicated, say it back in your own words and ask if you got it right. Use a closed question when you need one specific fact confirmed, and then ask for the explanation behind it. Poynter's advice on letting your questions guide the conversation covers the difference between open and closed questions and why double-barreled ones cause trouble.
You're done when you have at least one usable, attributable explanation of the process, the who, what, when, and scope, specific examples where the source has them, and a list of claims that still need evidence. If the source doesn't have a great anecdote, don't push for one.
Common mistake: asking two questions at once, so the person answers only the easier half. Long lead-ins, leading questions, and interrupting also cost you detail. NPR host Rachel Martin recommends patience and leaving room for silence after an important disclosure, though her advice comes from personal interviews, so use it with care and don't use silence to pressure anyone.
Pro tip: when something surprising comes up, put a star and a timestamp next to it, and follow up right away with "How do you know?" and "What would change your answer?" A surprising claim is worth more to you than a nicely phrased sentence.
Step 6: Close the interview by collecting the evidence and missing voices
The last few minutes of the call are where a lot of the value is. Recap the factual points you plan to check. Confirm the spelling of names and specialized terms. Ask for the original document, screenshot, methodology, or record behind any concrete claim, along with permission to use anything they send. Then ask who else you should talk to, including someone who had a different outcome, and how you can reach the person for follow-up questions.
If the source gave you something on background, ask whether one particular fact can go on the record. And if your article criticizes a person or company in a meaningful way, try to get their response and report honestly whether you received one. A marketing article doesn't have to follow newsroom rules to the letter, but the idea of fairness carries over in proportion.
You're done when the promised evidence, open questions, contact permissions, and any requested responses are logged and each has an owner. A note that says "waiting for documentation" is an open item and not a publishable fact.
Common mistake: hanging up right after the apparent quote, then finding out during editing that a result has no denominator or that the person can't share where the number came from.
Step 7: Verify, attribute, and place the reporting in the article
Now go back through your recording and notes and check every quote and paraphrase you plan to use. Trace each statistic to whoever originally produced it, and check the period, the population, the unit, the definition, and the comparison point. When two answers contradict each other, settle it with the source or an independent record, and don't just pick the one that makes the better headline.
Then go through the draft sentence by sentence and decide what each one rests on. It might be an original document, something an interviewee said, the company's own claim, or your clearly labeled inference. Name sources by their relevant role. Start each section with a short direct answer, and follow it with the specific example and its evidence. This structure also fits the way people read and the way AI engines pull answers from a page, and you can read more about it in DeepSmith's guide to writing answer-first content.
Here is a quick illustration of what this does to a weak sentence. This is an example of the method and not a customer result.
- Unreported draft: "The new process dramatically improves content quality."
- Reporting question: "Which part of the process changed, what happened in a particular article, and how did you assess quality?"
- Publishable version: describe the specific process and attribute the participant's observation. Include a measured improvement only if the measure can be checked, and otherwise leave the number or the cause out.
You're done when a reviewer can tell where each important claim came from and how solid it is without guessing. Quoted words should match the recording, paraphrases should keep the meaning, and the article shouldn't turn one person's experience into a rule for everyone.
Common mistake: stitching fragments together into a quote the speaker never said, quietly fixing a real error inside quotation marks, or attaching a number to an interview when the number has no support. The safest rule for a guide like this is to keep direct quotes word for word and to paraphrase awkward speech accurately, without polishing it inside the quotation marks. If a quote is too unclear to use even as a paraphrase, leave it out.
This is also the step where the writing itself gets produced, and DeepSmith can help there. Content Studio researches and produces brand-grounded articles with SEO and AEO structure, links, and metadata already included, and it has an editing and review stage. Your team still has to supply and check the facts that come out of your interviews, because DeepSmith doesn't conduct interviews or confirm what a source told you. What it can do is take a checked reporting brief and turn it into a reviewed draft in your brand voice.
Step 8: Run a reporting-focused prepublication check
Give an editor, or another reader who will look at it critically, the claim grid alongside the draft. Have them check names, titles, dates, numbers, units, quotes against the recording, the conditions you agreed with each source, and permissions for any documents or images. Ask them to challenge every "first," "only," "always," and "best," every claim that one thing caused another, and every result with no qualifier. If a fact can't be supported, either get the evidence, soften it so it says only what you know, or take it out.
Also decide ahead of time how you'll handle it if a reader or a source points out a mistake after publishing. The Society of Professional Journalists' code of ethics calls for prompt and prominent corrections, and that's a good habit for a company blog as well.
You're done when each high-stakes claim has a recorded verification decision, no quote contradicts the recording, and the editor knows what is still unconfirmed.
Common mistake: assuming a polished, SEO-ready article is a fact-checked one. Those are separate jobs. If you want a wider checklist for the later stages, DeepSmith's posts on fact-checking an AI draft and on human review and QA gates go into more detail.
Working with the interview format
People use "interview format blog post" to mean two different things, and it's worth knowing which one you're writing.
A reported how-to, like this one, combines checked interviews, records, and attributed examples into steps. The interviews are the raw material and the reader sees the result. An interview-format blog post is an edited question-and-answer exchange, where the reader sees the conversation itself. If you write an interview format blog post as a long Q&A, work from a full transcript so you don't pick answers out of context, edit for readability without changing what the person meant, and say so if you edited in a meaningful way. In either format, a source approving the draft doesn't replace your own accuracy check as the editor. A source can fix a misspelled name or a wrong number in a follow-up, but they shouldn't get control of the text or be able to swap out an accurate answer they now wish they'd phrased differently.
If you're inviting someone to be interviewed, a short note helps. State the reader's question, why this person has firsthand knowledge, where it will be published, the rough topics, and the format of the interview. Mention that you may ask for documents or follow up on facts. Don't promise favorable coverage or editorial control.
Pulling it together
The eight steps are journalism skills for content marketers packaged as a loop you can run on any post. Pick the claims that need proof, find people who know firsthand, prepare a flexible route, agree on the rules, ask for examples, collect the evidence, verify and attribute, and check once more before you publish. A few smaller points are worth keeping in mind as you do.
- Reporting techniques for content writing don't come with a set number of interviews or quotes a post needs, and it depends on the claims, the stakes, and whether your evidence is independent. Two or more sources is better than one where you can manage it, and that's a preference and not a quota.
- A recording shows what was said, not whether it was true. You still have to check it.
- Public officials, vulnerable interviewees, and employees who could face retaliation raise different ethical concerns, and anonymity shouldn't become a convenience. Don't promise confidentiality you can't protect.
- Google says a page has to be indexed and eligible for a Search snippet to show as a supporting link in AI Overviews or AI Mode, and that there are no extra technical requirements. Being eligible doesn't guarantee a citation, traffic, or a stable position, so it's better to publish a clear, accessible page and then watch what happens. DeepSmith tracks prompt-level and page-level citations, which lets you see results after publication, but a change in citations doesn't prove the interview caused it.
What to do next
Choose one upcoming post and find the claim in it that has the least support. Then set up one interview built to test that claim, using the question route from Step 3. Once you've verified what you learned, you can turn the reporting brief into a reviewed, brand-grounded draft. If you'd like to try that with your own team, you can start a free DeepSmith trial and see how Content Studio handles a brief, and you can set up prompt tracking to watch how the post does in AI search once it's live.
