Why this signal matters for Google
An interview is the least copyable content on your site. Anyone can rewrite a tips article in an hour. Nobody can rewrite an hour-long conversation with the quality manager of a cheese dairy. The sentences are hers, they were said on a Tuesday in March, and you hold the audio file. It belongs to the same family of proof as a lived testimonial, with one useful difference. Here the proof does not rest on your word. It rests on someone else's, and that person has a name, a job and a reputation to protect.
Google has human evaluators grade pages, and this is exactly the kind of trace they look for. Those evaluators work from a guide Google publishes and anyone can read, the Quality Rater Guidelines. Section 3.4 asks them to check whether the person publishing really knows the topic. Section 7.3 also admits knowledge drawn from practice: what someone who does the thing every day knows, diploma or not. An interview ticks both boxes at once. It shows you know what to ask, and it brings into your page the experience of someone who lives the subject daily. Section 3.3.4 points the same way: it counts expert co-authors and citations as evidence of a good reputation.
An interview only counts if the reader can check it. Three elements make the difference, and they take five minutes to add. The person's full name, with their role and a link to a public profile: "an industry expert told us" proves nothing at all. The date of the conversation, because a read on a market ages in two years. And the recording, or at minimum a full transcript, which shows you did not write the answers yourself. It is the same standard you apply to your cited sources: you do not ask the reader to take your word, you hand them the means to verify. In the interviews I proofread before they go live, the detail missing most often is not the guest's name: it is the date of the conversation, and without it the reader cannot tell whether the opinion is two months or five years old.
On sensitive topics, an interview changes status. If your page touches health, money, law or safety (Google calls these topics YMYL, for Your Money or Your Life), section 3.4.1 tells raters to look for professional expertise, not lived experience alone. Interviewing a named, verifiable pharmacist about drug interactions does not replace your own competence, but it ties your page to a person who has it. For AI systems the effect is direct. When ChatGPT, Perplexity or AI Overviews decide what to cite, they look for what exists in one place only. A sentence spoken by an identified person, published on your site and nowhere else, is precisely that.
The NeuroEEAT scoring grid (0-10)
Rate what your interview proves to someone who does not know you and spends two minutes checking.
| Score | Description | Concrete indicators |
|---|---|---|
| 0-2 | No interview at all | No content gives the floor to anyone but you. Everything is written secondhand, from what already exists online. |
| 3-4 | Unsourced quote | Sentences credited to "an expert" or "an industry professional", with no name, no role, no date. There is no way to tell whether the exchange happened. |
| 5-6 | Named interview | The person is named, with their role. The questions stay generic and appear in ten other articles. No recording, no interview date. |
| 7-8 | Documented interview | Name, role, link to a public profile, date of the conversation, questions specific to the topic and a full transcript published. Anyone can reread what was said. |
| 9-10 | Proven interview | Audio or video on the page, transcript, a photo taken during the exchange, guest marked up in the code, and answers found nowhere else. The guest shares the page from their own account. |
How to audit it on your site
Are your interviews actually working for you? Start with the pages that get the most visits.
- 01
Count the voices that are not yours
List your twenty most important pages and mark the ones where an outsider truly speaks. If none carries words you gathered yourself, your site only has your own point of view to offer.
- 02
Check that every person quoted is identifiable
For each quote, look for the full name, the role and a link to a public profile. QRG section 3.3.4 rewards identified experts. An anonymous quote cannot be verified, so it does not count.
- 03
Look for the date of the conversation
The publication date is not enough: an article republished in 2026 can carry an exchange from 2021. State the month and year the conversation took place, separately from the page update date.
- 04
Test whether your questions are reusable
Read your questions again and ask whether they could be put to anyone in the trade. "What are the trends in your industry?" is an empty question. "Why did you drop hand cutting in 2024?" belongs to one person only.
- 05
Find the raw material
For every published interview, confirm the recording or transcript still exists and that you know where it is. Without raw material you can neither prove the exchange nor fix it if the guest disputes a sentence.
How to activate or improve it
From a first email exchange to an interview other people cite.
Let a real person speak
- Pick someone who does the work every day rather than the best-known name: a shift lead, a supplier, a long-standing client.
- Send five questions by email and ask for written answers, with permission to publish them.
- Publish the full name, role, company and date of the exchange at the top of the page.
- Keep the original email in a dedicated folder: it is your proof if anyone disputes a sentence.
Record it and document it
- Move to a live exchange, by phone or video call, and record it after asking for consent out loud.
- Prepare ten questions built on the person's actual work, not on the articles already published about the topic.
- Publish the full transcript below the summary, flagging any cuts made for clarity.
- Add a link to the guest's public profile and a photo taken during the exchange.
- Send the page to the guest before publishing: a correction beats a retraction.
Turn it into a citable resource
- Put the audio or video on the page itself, with a timestamped transcript right underneath.
- Mark up the guest in JSON-LD (a structured description written in the page code and read by search engines) with a Person object and a sameAs property pointing to their professional profile.
- Pull two or three dated or numbered statements out of the interview that nobody else publishes, and surface them in the page.
- Build a series: four interviews in the same niche make your site the place where the trade speaks.
- Write down who on your side keeps the recordings, and for how long.
5 common mistakes to avoid
Sector specifics
Let qualified people speak
On a page about treatments or symptoms, section 3.4.1 wants professional expertise behind the text. Interviewing a named health professional, with their licence number or a link to an official register, ties the page to someone qualified. State plainly what the interview does not replace: a consultation.
The trade next door as guest
A roofer who publishes a conversation with the carpenter he has worked with for ten years creates content nobody else in town can copy. The name, the town and the year they started working together anchor the page in a place, which local search rewards.
Interview your own client
In B2B, a client interview is the spoken version of a case study. Have the person who lived it describe the problem, the decision and the measured result. A written agreement on the figures you publish prevents a takedown six months later.
Go back to the maker
Interviewing whoever builds or repairs the product surfaces details no supplier sheet contains: why that material, which part fails first, what maintenance doubles its life. Those are the sentences search engines and AI answers pick up.
Signals to activate in parallel
An interview gets stronger when it leans on the other signals that prove who spoke and what was said:
FAQ
How do you conduct an interview when you know nobody in your field?
Start with your first circle: a supplier, a loyal client, a former colleague. An interview with a tradesperson you have known for five years beats a refusal from a famous name. Ask by email, explain the topic, promise five questions and a review before publication. The response rate is far higher than most people expect.
Should you publish the whole interview or only the best parts?
Publish a structured summary at the top and the full transcript below. The summary serves the reader in a hurry, the transcript serves the proof. If you cut, say so with a line such as "answers condensed for clarity". The one thing never to do is rewrite the guest's sentences while keeping the quotation marks.
Does an email interview count as much as a live conversation?
It counts, but less. Written answers tend to be smoother and yield fewer unexpected details. They still work as proof if the name, role and date are published. A live exchange brings the hesitations, the examples and the numbers a prepared answer leaves out.
Do you need written consent from the person interviewed?
Yes, and it takes two lines. Ask in writing for permission to record and publish, and let them review the final version. That protects the relationship as much as the page: a guest who approved their words shares them, while a surprised guest asks for a takedown.
How many interviews does it take for the signal to count?
One properly documented interview already changes what a page is. For the signal to weigh at site level, aim for a series: four to six interviews in the same niche, published over a year, are enough to make your site the place where people in the trade speak.
Case study
Before · one more article on the topic
After · words nobody else has
Illustrative figures meant to show the typical progression of the signal. They do not come from a real client case and must be replaced with your own measurements before any external communication.