We’re rebuilding the free AI Visibility Snapshot — the unglamorous part matters
Quick build-in-public update 🙂
We’re making a substantial improvement to the free GEO Tracker AI Snapshot / Grader.
Early internal results from the new version look very promising, but we’re finishing the last QA before opening it to everyone. The goal is not to publish another generic “AI visibility score” — it is to make the result more useful, more explainable and more honest about what was actually measured.
A few of the things we are tightening right now:
buyer questions now have to match a real measurement intent: discovery, use case, evaluation, comparison or desired outcome;
localized prompts are being refined for Czech, Slovak, German, French, Spanish and Polish — not just translated word-for-word;
report headlines, evidence labels, recommended actions and timestamps are being presented natively for each supported market;
observed competitors, directories, citations and website evidence need to stay distinct. If they are mixed together, scores and recommendations become misleading very quickly.
This is the less glamorous side of GEO/AEO/AI SEO: a market and language change is not simply a dropdown. It changes the buyer questions, the AI answers and the evidence that has to be interpreted.
We’re in the final stretch now. Once the checks are complete — hopefully within the next few hours — we’ll publish an update and invite everyone to try the improved free Snapshot.
We’ll share what changed, not just say “new version shipped” 🚀
Comments (8)
We’re fine-tuning the final details. After that, everything should be fully functional and ready to use.
Someone is on fire to make GEO Tracker the best product out there 👍🏻🙂
Haha, guilty as charged! 🔥🙂
Right now the “on fire” part is mostly coffee, test cases, and prompts behaving differently across six languages. The goal is to turn all those weird edge cases into boring, reliable results — which is probably the least glamorous definition of product love.
Thanks Sergey! 🚀
↳ Replying to Petr GEO Tracker AI
I have full confidence you will make it work, Petr! I can see from your activity here and how your product works already that you have the right energy to make it a success. Keep going! You are clearly on the right track.
↳ Replying to Sergey Kargopolov
Thank you, Sergey — that genuinely means a lot, especially coming from someone who has already built and learned from real projects.
I have a very positive feeling about SaaS Hive already. It has a rare, almost family-like atmosphere: people are genuinely kind, the community is growing, and founders seem willing to share what is actually working and not working.
I want to be part of that beyond GEO Tracker AI itself — to learn from other founders’ challenges and experience, and hopefully help where my own experience is useful too. I have made plenty of mistakes over the years: bad decisions, wrong application architecture, imperfect business plans and ICP assumptions. Those lessons are often more valuable than the successes.
I have also worked in sales for more than 20 years, from sales representative roles to more solution-oriented sales work, so I hope I can contribute on both the technical and commercial side when it is helpful.
GEO Tracker AI is no longer a side project for me — it is my full-time focus now. And SaaS Hive already feels like a valuable place for both professional and business growth.
This is the kind of update that makes the product feel more credible to me than simply announcing another scoring feature. The distinction between competitors, directories, citations, and first-party website evidence seems especially important because combining them would make the recommendations look more precise than they really are.
I also like the decision to localize buyer intent rather than just translate prompts. A comparison query in Germany or France may not behave the same way as an English-language query, even if the literal wording is equivalent.
The part I’d be most interested in seeing when this launches is the explanation behind the score. If the Snapshot can clearly show what was measured, what evidence was observed, what remains uncertain, and what I should do next, that would make the free grader much more useful than a simple lead-generation score.
Four things I’d be curious to test: how much results differ between markets for the same company, how you handle inconsistent AI answers across repeated runs, whether recommendations change meaningfully based on query intent, and how much of the free Snapshot is meant to stand on its own versus lead into the paid tracker.
Thank you Jinny,
these are exactly the questions we want the methodology to answer transparently.
For market differences, we treat the market and language as part of the measurement context, not as a translation setting. They affect how buyer questions are generated, how the AI engine is queried and which competitors or sources are considered relevant. A German and a French measurement for the same company are therefore two separate measurement series, not directly comparable points in one trend.
For inconsistent AI answers, our standard paid monitoring cadence is weekly. On Pro, each monitored prompt is measured three times per AI engine within the same weekly run. We then aggregate those readings into a mention rate and confidence level instead of reducing the result to a single binary “mentioned/not mentioned” answer.
Three repetitions do not eliminate uncertainty or prove a permanent result, but they help us distinguish consistent visibility from a one-off answer. A result appearing in three out of three readings is different from one appearing in only one out of three.
Pro currently measures across ChatGPT, Perplexity, Google AI Mode and Gemini. A limited number of Hero Prompts can be monitored daily, but those use one reading and are treated as a faster trend signal, not as the same confidence signal as the weekly repeated measurement.
Query intent is also an important part of the methodology. We separate buyer questions into discovery, use case, evaluation, comparison and desired-outcome situations. Recommendations should then be tied to the actual question that was lost, the competitor that appeared and the source evidence observed. If every intent produces the same generic recommendation, we consider that a product failure rather than a useful action layer.
The Free Snapshot and paid tracker have different jobs. The current English Snapshot is intentionally a one-time, cost-bounded baseline: five buyer-intent questions measured through one Perplexity run. It should still stand on its own by showing what was measured, what was observed, the limitations of the sample and the most credible next action. If it only displays a low score and hides the explanation behind the paid product, then we have failed.
The paid tracker adds the larger prompt universe, more AI engines, repeated measurements, weekly history, competitor tracking and the ability to see whether results move after an action. Our Outcome Loop records a completed action and compares later measurements over a fixed 14-day window. Even there, we distinguish directional movement from proven causation and allow the honest result to remain inconclusive.
So the short version is: Free should answer “What does the current evidence suggest, and what should I investigate first?” Paid monitoring should answer “What is changing over time, how consistently, and what should I work on next?”
Thank you so much for such thoughtful and detailed feedback. It’s clear you looked beyond individual features and focused on whether the product can genuinely help a founder make better decisions and give them a reason to come back. That is exactly the question we are trying to answer with GEO Tracker AI.
You are right that the strongest part should not be the score itself. The real value is the path from a buyer question, to the AI answer, to the competitors being recommended, to the sources influencing that answer, and finally to a useful next action. We intentionally separate competitors from citation sources, because otherwise it becomes too easy to confuse who AI recommends with what AI is relying on.
The onboarding flow you described — domain, company and buyer context, suggested buyer questions, and founder review before measurement — is already part of the product. We are still improving the quality of those questions across different business types, because we do not want prompts that are only technically valid; they need to reflect real buyer intent.
Your point about the landing page is especially fair. Transparency around methodology and limitations matters to us, but it should not come before the immediate “aha” moment. I will work on applying this insight as quickly as possible: showing a simple, concrete example first, then letting people explore the methodology afterward.
I also agree that there can be too many concepts and metrics for a new user at once. We are improving the reporting experience so that mentions, recommendations, and citations are easier to understand, but the bigger goal is to make the product more opinionated: what is the biggest visibility issue, why does it seem to be happening, and what should the founder do next?
Your feedback on recommendations is probably the most important part. We do not want to stop at generic advice such as “improve comparison content.” The goal is to connect each recommendation to the relevant buyer prompts, competitors, sources, and evidence, then propose the most credible next step. We are still working on making this consistently specific and useful.
The three weekly questions you suggested — what changed, why did it change, and what should I work on next — are a very strong product direction. The same is true for alerts around a new competitor, losing visibility for an important buyer question, or gaining a recommendation where the product was previously absent. I will be working to incorporate these insights as quickly as I can.
I also agree with your view on pricing. At $129/month, the product needs to save meaningful research and decision-making time, or show that AI visibility matters for a particular category. That is one reason we offer the free diagnostic first: founders should be able to understand whether this is a meaningful problem for their business before paying.
Connecting AI visibility to business impact through Search Console, referral traffic, conversions, or CRM data is also an important direction for us. We want to make that connection more useful without pretending that every visibility change has a perfectly provable commercial cause.
Thank you again. Your final point is exactly right: the real test is whether the recommendations remain useful after a few weeks of tracking a real product. I would be very happy if you stayed in the program and continued testing it. If you do not mind, please send me a quick reminder shortly before your 14-day trial ends, and I will make sure you receive a very nice thank-you gift as well.
I will share incremental updates as they go live, both in our changelog and on SaaS Hive.
Sign in to comment or upvote.