A year ago, a growing share of the questions people used to type into a search box started getting answered somewhere else: inside ChatGPT, under a Google AI Overview, in an AI Mode panel, in a Perplexity thread. The link, the blue result, the click you optimized for two decades, quietly stopped being the whole story. We started LLM Pulse because that shift.
Twelve months in, we want to do something different from the usual anniversary post. Instead of a highlight reel, here is an honest look at what we believed when we launched, what actually happened, and the lessons we are carrying into year two.
Table of Contents
The bet we made a year ago
The premise was simple: if answer engines are becoming the place where people research products, brands, and decisions, then being visible inside those answers matters as much as ranking on a results page ever did. That visibility needed a name, a set of metrics, and a tool to track it.
Our first swing was an early MVP, a ChatGPT tracker that answered one question: when people ask ChatGPT about your category, does your brand come up? From there we built for AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization): share of voice inside AI answers, which sources get cited, how your brand is described, and how often you show up across the models that matter.
We were three founders, fully bootstrapped, in direct contact with every customer. That constraint turned out to be the best thing about year one.
What changed in AI Search in twelve months
The market did not sit still while we built. AI Overviews and AI Mode pushed generative answers in front of a huge slice of everyday Google queries. ChatGPT search and shopping matured. Perplexity kept growing as a research habit, not a novelty. New models, new answer surfaces, and new citation behaviors landed on a near-weekly basis.
“For twenty years the job was ranking a page and counting clicks. That playbook does not disappear, but it is no longer the whole scoreboard. Share of voice in an answer, which sources get cited, how a model describes you: these are the metrics teams now have to own. What I am proudest of this year is helping people see that AI Search is not a threat to SEO, it is its next chapter, and giving them the numbers to work in it with intent.
The practical takeaway for brands: your visibility is now split across many engines that each read the web differently, cite different sources, and describe your brand in their own words. Optimizing for one is not the same as optimizing for all of them, and none of it holds still long enough to set and forget.
5 things the first year taught us
1. Monitoring and optimization are one loop, not two products
Early on it was tempting to treat tracking and improving as separate jobs. In practice, the teams who win treat them as a single loop: measure where you stand across engines, act on what the data says, then measure again. That is why LLM Pulse ships both sides in every plan rather than selling visibility as a dashboard you stare at.
2. Model coverage is not a feature, it is the whole point
A brand can be a top citation in one engine and invisible in another. Tracking a single model gives you a comforting number and a misleading one. We made five base models part of every plan (ChatGPT, Gemini, Google AI Mode, Google AI Overviews, and Perplexity), with more available as add-ons, precisely because the gaps between engines are where the real work lives.
3. Language and geography are table stakes, not upsells
AI answers are local. The same query returns different brands and different sources depending on country and language. We decided early to track every country and language in the world at no extra cost, because charging per market quietly punishes the businesses that most need to see the full picture.
4. The fastest feature is the one a customer asked for today / now
Being small and bootstrapped meant we could talk to the people using the product every day and, often, ship their feedback as a live feature within hours. Agent and traffic analytics, GEO testing, GEO Writer, Reddit Intelligence, and the Zapier and n8n connectors all trace back to real conversations, not a roadmap written in a vacuum.
5. AI Search rewards a habit, not a launch
The most common mistake we saw was treating AI visibility as a one-off project. Answers move constantly as models update and as competitors, publishers, and third-party sources reshape what gets cited. The brands that gain ground are the ones that monitor on a cadence, test deliberately, and iterate. AI visibility is a discipline, not a one-time fix.
Year one, in review
| What we assumed at launch | What the year taught us |
|---|---|
| Brands would want a monitoring dashboard first | They want to act on the data, so optimization has to sit right next to it |
| Tracking two or three models would be enough | Coverage across many engines is where the surprises, and the wins, hide |
| Enterprises would move first | Teams of every size moved, across banking, health, insurance, marketplaces, SaaS, and consumer brands, plus agencies |
| A quarterly roadmap would guide us | Direct customer feedback, shipped fast, guided us better |
Along the way we also shipped things we are proud of: prompt tracking, citations, sentiment, share of voice, model comparison, GEO Optimization and GEO Testing, Content Intelligence, Prompt Research, owned media tracking (your website isn’t your brand’s only asset.), an API, MCP and CLI, a Looker Studio connector, a Chrome extension, white label for agencies, and an embed version for SaaS. We also launched Tremor, a free weekly AI Search volatility tracker, so anyone can see how much the answer engines moved that week.
The best trivia did not fit here. We collected the milestones, the surprising numbers, and a few behind-the-scenes moments from year one in this LinkedIn post.
Where we are heading in year two
The direction has not changed, but the stakes have. AI Search is moving from an experiment on the edge of marketing to a core channel that teams are held accountable for. Our job is to keep it measurable, keep it fair across engines, and keep the loop between seeing and improving as short as possible. We are still bootstrapped, still talking to every customer, and still shipping fast.
Thank you to everyone who trusted a one-year-old product with something as important as how their brand shows up in AI. You shaped this more than any roadmap could have.
“A year ago we were explaining what GEO even meant. Now teams open the tool to see how they show up across engines and, more importantly, what to do about it. The best part of bootstrapping is that when someone asks for something on a call, we can often build it before the call ends. That is the company I wanted, and it is the one we get to keep building.
Try LLM Pulse
If you want to see where your brand stands across ChatGPT, Gemini, Google AI Mode, Google AI Overviews, and Perplexity, you can start in minutes. Plans begin at €49/mo, and there is a 14-day free trial with no mandatory sales call. Track every country and language, monitor and optimize in the same place, and turn AI visibility into a habit rather than a guess.
Measure it, improve it, and build a brand that AI recommends.
Happy birthday LLM Pulse.
