Interview with Ken Marshall, Co-Founder, Meet Sona

This interview is with Ken Marshall, Co-Founder, Meet Sona.

What was the pivotal moment that set you on the path to building and scaling in AI-driven SEO and SaaS?

I had built an elaborate system for my LinkedIn content consisting of Google Sheets, ChatGPT voice mode, a customized GPT trained on my style, and social media repurposing and automation tools. However, one day, when I was looking into n8n for further automation potential, I decided that it would be easier to build a product that did everything I wanted.

AI made sense since it was already being used and could eliminate the friction I was experiencing to get the outputs I wanted.

Building on your success with original trend reports, if you had 30 days to ship one from scratch, what exact process would you follow?

The easiest way would be to leverage my LinkedIn audience. They are folks who are mostly in my Ideal Customer Profile (ICP), and they are quite engaged. Rather than paying for a survey or doing a meta-analysis of existing bodies of work, I would go directly to the people in my audience who represent the customers for the product, ensuring there is some overlap.

I would include a balanced mix of qualitative and quantitative questions, probably keeping it to six to eight questions total.

I would reward participants somehow; maybe with an Amazon gift card or a set of credits for the product, but I would incentivize them to complete it.

On the back end, I would use Claude Opus or hire a virtual assistant (VA) to gather other sources. We would try to find the gaps in existing research around trends, as this allows us to manipulate the data or present it in a way that offers built-in information gain when we publish it.

This approach helps ensure that we are cited more frequently by reputable sources, which is beneficial for building links and domain authority. Additionally, it will be genuinely useful when we begin to promote it. For example, we could say, "Hey, did you know these three reports are great for this topic? And our report is excellent for this other topic as well?"

As you’ve adapted for answer engines like ChatGPT and Perplexity, what is one change to your content calendar that produced measurable results?

The biggest change we've made so far is leaning more heavily into really bottom-of-funnel and middle-of-funnel assets. This includes:

Case studies
Deeper information on pricing pages
Competitor and versus pages

These are areas where we can add valuable information while also shaping the sentiment of the brand and how we want to be perceived.

In addition, we've also focused more on digital PR with a more editorial process. In the past, it was possible to use tactics that were weighted more favorably in Google Search and Bing Search. However, it's now crucial to form legitimate partnerships. For instance, if there’s someone who has a roundup of the best industry software, forming a partnership with them on LinkedIn, through outreach, or by inviting them on your podcast can be significantly more valuable. This approach has been beneficial for a few years, but especially now, as the models are heavily biased toward those types of content assets. The ratio is about 90-10, with the 10 being everything else. Therefore, really getting granular with those kinds of assets has been our main focus, alongside high editorial quality, digital PR, and outreach.

Another important aspect is being more explicit with elements like FAQs. These topics should have appeared in the content anyway, but we're now being more intentional about answering them in a manner that does not compromise the user experience. This could mean placing answers at the end of the article or throughout, but in an accordion format that uses HTML and CSS, ensuring it is crawlable and understandable.

Lastly, we have refrained from branching out into areas where we do not have deep topical authority. While we may have experimented with this in the past, we are now focused on sticking to our core topics in which we hold expertise.

On Reddit, LinkedIn, and Medium, what posting cadence and post formats have directly led to attributable pipeline for you?

I'm going to be completely honest here. LinkedIn is where we've seen 90% of our attributable pipeline. Reddit and Medium? We've experimented, but they haven't moved the needle for us in any meaningful way.

On LinkedIn, I post 3-5 times per week. And I'm not just talking about polishing up some AI-generated thought leadership content. I'm talking about real, unfiltered takes on what's working in our business, what we're learning about founder-led content, and what I'm seeing in the market. The formats that have directly led to pipeline for us are pretty straightforward:

Personal stories with a lesson attached. These perform incredibly well because people can feel the authenticity. For example, when I talked about our trial conversion challenges or when we hit our first revenue milestone. Those posts get engagement; more importantly, they start conversations in DMs.
Tactical breakdowns of how we do something specific. Not surface-level "here's 5 tips" content, but actual walkthrough-style posts where I show our process. When I break down how we structure our AI interviews or how we think about verbal identity, those posts attract the exact people who need what we're building.
The contrarian take format. Posts where I push back on the "AI will do everything for you" narrative or talk about why most founder content sounds soulless. These posts get shared more and position us differently in the market.

Here's what I've learned about cadence: consistency beats frequency. I used to stress about posting every single day, but what actually matters is showing up regularly with something worth saying. Three great posts per week beat seven mediocre ones every time.

The other thing is that I engage way more than I post. I'm in the comments, starting conversations, and DMing people who resonate with what we're doing. That's where a ton of our pipeline actually comes from; not the post itself, but the conversation it starts.

As for Reddit and Medium, I'm not saying they can't work. They just haven't been worth the time investment for us given our ICP and where they spend their time. Your mileage may vary depending on your audience.

For an early-stage AI SaaS with limited resources, what is your minimal viable SEO stack across people, tools, and weekly rituals?

First off, people. You need one person who actually understands SEO strategy, not just the mechanics. This could be you as the founder if you're willing to invest the time to learn, or it could be a fractional SEO who works 10-15 hours a week. Do not hire a full-time SEO person yet unless you're already seeing traction. Then you need one solid writer who understands your space and can write content that doesn't sound like ChatGPT threw up on a page. That's it. Two people or roles.

For tools, I'm going to give you the bare minimum that actually matters. You need Google Search Console and Google Analytics, both free. That's your foundation for understanding what's working and what's not. Then you need one paid SEO tool. I'd go with Ahrefs or SEMrush. Pick one; it doesn't matter which. You're looking at maybe $100-200 a month. Use it for keyword research, competitor analysis, and tracking rankings. That's it. Don't get distracted by the 47 other features.

You also need a way to manage your content calendar. This can literally be a Google Sheet. I'm not joking. We used a Google Sheet for our first year, and it worked perfectly fine. Then you need some kind of content management system. If you're on WordPress, great. If you're on Webflow or Framer, also great. Just make sure you can actually control your meta tags and site structure.

Now here's where most people mess this up. They think they need heat mapping tools, rank tracking dashboards, and all this other stuff. You don't. Not yet. Get the basics right first.

Weekly rituals are where the magic actually happens. Every Monday, spend 30 minutes reviewing your Search Console data from the previous week. What queries are you showing up for? What's your CTR? Where are you on page two that you could push to page one with some updates?

Every week, publish one piece of really solid content. Not five mediocre posts—one great one. Bottom of funnel stuff first: solutions pages, comparison pages, use case pages. The stuff that actually converts. You can worry about top of funnel educational content later once you have money coming in the door.

Then, every week, spend time updating one existing piece of content. This is huge, and most people skip it. Find a page that's ranking in positions 5-15 for something valuable and make it better. Add more depth, update the stats, improve the examples. This often gives you faster results than creating new content.

To deliver the “information gain” LLMs want to cite, where on a page do you inject unique data or methodology so models and editors prefer you?

Here's the thing that most people get wrong about this. They think information gain is about stuffing some proprietary data into a random section and calling it a day. That's not how this works.

The best place to inject your unique data is right after you establish the problem or common approach. Your structure looks like this: introduce the topic, acknowledge what everyone else is saying, then boom, hit them with your unique take or data that shows a different angle.

Place that insight early, ideally in the first third of your article. The models scan content and weight information that appears earlier more heavily. If your unique data is buried at the bottom, it's less likely to be picked up.

Now here's where people mess this up. They create elaborate methodologies but don't explain them clearly. After you present your unique insight, immediately follow with a methodology section. Two to three sentences explaining your sample size, time frame, and approach are essential. Something like "We analyzed 10,000 cold emails sent between January and March 2024 by early-stage B2B SaaS founders. We tracked open rates, reply rates, and meeting bookings across five industries."

The other place to inject unique information is in case studies or examples. Pull from your own experience or your customers' results. The models are looking for real-world application, not just theory. Show actual data. For example, "One of our customers went from a 2% trial-to-paid conversion to 8% after implementing a weekly founder interview cadence."

Visual data presentation matters for getting cited. If you have unique data, present it in a simple table or clean chart. The models parse that information more easily.

The models prefer recent data over old benchmarks. If you can say "based on data from Q4 2024," that signals your information is current and relevant.

One more tactical thing: If you have a unique framework or process, name it. The models are more likely to cite something with a clear name attached to it. For instance, "the Verbal Identity Engine" or "the Information Gain Framework."

Focus on creating information that actually advances the conversation. Not "what will game the algorithm" but "what would help someone understand this better than anything else out there."

Thinking of a campaign that underperformed, what did you change in timing, positioning, or distribution that you now use by default?

One of the campaigns we launched that didn't perform well was targeting a segment with a new way that folks described the work we did. We realized that a lot of VPs of marketing at B2B SaaS companies don't really describe the work they need as organic growth marketing. In reality, other lower-level marketers, the ones who weren't making decisions, were using that term.

Once we created a new hub and spoke for the term organic growth marketing, we tried to reposition our homepage accordingly. While we did gain some visibility and website visitors, the conversion rate was not as good as for phrases people actually understand, such as modifiers like SaaS SEO agency and B2B SEO firm.

The lesson learned was that we took the risk of trying to expand our footprint from an organic search standpoint, but we didn't start by examining form submission language or engaging with people on LinkedIn regarding their observations, or talking to our customers. Now, we aim to work backward from customer needs, expectations, and jobs to be done, as well as the messaging and language they use, rather than forcing a connection from identifying a problem to finding a solution.

The big takeaway is to always validate your positioning with actual customers before you go all-in on a campaign. We could have saved ourselves weeks of work if we'd just asked five VPs of marketing how they describe what they need. Now, that's our default: customer language first, SEO opportunity second.