So… What does an SEO Product Manager do in 2026?
Back in 2023, I wrote a blog post with this same title. It was time to update it. Product Managers working in SEO are still a niche within our (already niche) industry, and I imagine the experience of PMs varies widely from company to company.
For instance, I have my own engineering teams. It’s a luxury not all SEO PMs have. I know of others who still have another layer of PMs between them and the SWEs (Software Engineers). It comes with its own challenges, such as making sure nine engineers have enough relevant work to do. AI has made them much faster, which forces me to step up too. There isn’t a single quiet day.
This blog post is an updated account of my personal experience in the role of a Product Manager. Four years later, it’s still the most rewarding job I’ve ever had. It still gets me out of bed excited every morning and motivates me to write this blog post on a Sunday at 7 am.
Here we go: what does an SEO Product Manager do in 2026?
Choosing what to fix
When I took this role back in 2022, I thought my job would be to fix technical SEO issues. Improve page speed, implement structured data, fix broken links. I finally didn’t have to beg for a few developer hours for SEO, so I could get everything done.
While I did have the manpower to do all these things, I quickly realised that my job wasn’t to fix things – it was to build them. Over time, you realise that you can keep finding issues to fix, things to improve, and things to adjust to Google’s latest guidelines or updates because a competitor did something. This is the wrong approach.
Fixing everything isn’t sexy, doesn’t give you visibility and, worst of all, often doesn’t have a big impact. In fact, people often aren’t even aware that something is broken until you point it out. So I had to choose which “fixing” doors to open. For example:
- Some pages are slow to load: is their business impact significant?
- A section of the website is getting no organic traffic: have AI Overviews taken most of the clicks anyway?
- There’s a high volume of broken links in a given market: is there enough business demand in that country to justify fixing them?
Once you realise there’s always something to fix and weigh the potential impact of one opportunity against another, you learn to accept that a few things will remain “broken”.
In 2026, I’m more adamant about completely ignoring low-impact work just to “tick a box” – you’ll see why further down. When I have enough doors open, I don’t even peek behind some of the others.
Pitching a roadmap
Before becoming a PM, I’d just do the next things that came to mind, usually whatever was easiest to implement on a website. Today, I couldn’t live without a roadmap.
Similar to 2023, I choose a few big things to work on every quarter. The difference is that, with AI speeding up SEO research and making engineers incredibly faster than before, each of the three or four big initiatives I choose to work on is bigger and bolder. Here’s an example of a roadmap I had for a quarter in the distant past (I couldn’t reveal the current one):
- EEAT Optimisation Program
- Bulk Internal Linking CMS Tool
- Build hub pages
- Bulk Title and Description Update CMS Tool
At that time, a feature like automated internal linking was big enough to count as a big bet. Today, a big bet is closer to redesigning the user experience. A great internal linking structure or key takeaways in a blog post are just table stakes.
My roadmap is still limited to three or four big ideas. Those bets still need to tell a good story to leadership and be easy to understand as a concept (even if they’re complex to build). But the research involved in going from an idea to code, identifying the teams involved and understanding the overall complexity has changed. Instead of digging through wikis and Slack channels, I can now ask MCP servers where to find the answers.
Before, I’d pitch an idea to my team to find out if something was possible. Now, I’m able to do that research myself and bring it to them at a much later stage.
Building prototypes
This is an area where my work as a PM has really changed. My design skills go as far as Canva, using Inspect in Chrome to change some text or cropping images in Google Slides. However, using AI (and, luckily, some internal MCPs that understand our design systems), I can build a working prototype. I used to picture these features; now I can ask AI to build them and keep iterating until they’re good enough to sell.
I still write a Product Requirements Document, aka PRD. Yes, I draft it with AI and then edit the whole document. With only a short amount of time to pitch these ideas to my skip-level manager, I proposed four ideas in 15 minutes this quarter by demoing my prototypes. These were built using Claude Code hooked into internal MCPs (although they also work without them, using Claude Code alone). It was much easier and more exciting to sell a testing experience than a Word document. I pitched four ideas, and we’re building three of them. Not bad.
While I quickly fell in love with these prototypes, my jaw dropped when I saw the actual designs proposed by UX. The concept was the same, but wow, the quality was on another level. AI took me far, but working with UX turned these prototypes into a much more polished experience, much more on-brand – and raised plenty of questions I hadn’t anticipated.
I was surprised that my UX counterpart was glad I’d given him a prototype to work with. Once these experiences are released, I’ll update this post to compare my prototype with the real product.
Information gain opportunities
If you do Enterprise SEO, there’s a good chance your company has plenty of rich data. I realised there was a lot of publicly available data I could reshape in different ways and combine with other data that, after getting the necessary authorisation, I could make public.
To give you some examples, we aggregate salary information and job demand, and extract skills and benefits from job postings. I own three products and have enriched all three with this data. This is one of my first starting points for making pages richer in content, ranking for more keywords, capturing more fan-outs and, most importantly, making users more satisfied when they land on my content.
It’s all about a good story
Even before my PM days, I learned a long time ago that you always need a good story in marketing and product. I once won two awards for a fairly standard migration. It just happened to take place during the busiest week of the year for the company, which made the risk much higher. The story was the when, not the what. The what was quite ordinary; the when was exciting.
I had clients who cared about one thing above all else every month. So, if I made good progress on that one thing, I could buy time for everything else. Back in my agency days, my director would look at NPS scores to measure how well I was doing my job. He didn’t live in the SEO biz; he lived in the business of keeping clients paying. And good NPS scores meant lower churn.
Now, I need to read the room – many rooms, across different orgs – as well as company-wide emails and hints from managers, to understand what is most valued.
For instance, if I tell my VP why Google dislikes AI content, he might not care much. He’s looking for a solution. If I say that we’ve found a way to draft content with AI, use our unique data to make it more valuable for the audience, and keep it fresh, then I’ve got his attention.
So when I build a roadmap and a prototype, I’m also selling a story of success. The semester has just started, nothing has actually been built, and perhaps the ceiling for my performance review has already been set. If my ideas are weak and don’t draw attention, it’s unlikely that even my best work will bring a high reward for the numbers we’re working towards.
However, if I set expectations high, the chances of people keeping this in mind throughout the semester are higher. You need a combination of strong results and a story that’s easy to follow and exciting to be part of to get the attention, trust and respect of the people above you.
How AI changed my day-to-day as a PM
I never believed AI would make things faster and, therefore, give us our personal time back. We do things much faster, yes, but because the expectation is to do things faster, we end up doing a lot more. As a Product Manager, I use AI and tons of MCPs in every part of my work.
These days, I can ask questions directly about the code, so instead of raising an investigation ticket and waiting a week, I can already do the investigation on my own.
- How can we filter this using taxonomy?
- What API is used to query this data?
- In which countries have we implemented this or that feature?
And in the end, I even ask AI to write a ticket with the technical implementation requirements. Heck, AI even creates the ticket on my Jira board. It’s wild.
Sometimes, an idea dies right there at the research stage, and I never waste someone else’s time. In most cases, my initial request is already two steps ahead by the time it reaches an engineer.
Yes, I work hard to avoid AI slop. My rule of thumb is that if I can’t explain what’s written, or if it’s so long that I wouldn’t read it myself, I need to make it better. I’d rather have an incomplete request and let the right person figure out the details than just toss trash their way.