Show HN: Gumshoe.ai – SEO for AI
21 comments
Interesting approach, I'd love to learn more about.
We have been testing different solutions for this problem, and for now Nightwatch's LLM tracking stands our particularly.
They allow you to configure the phrases manually, and have more control to set it up.
You might want to check for yourself: https://nightwatch.io/ai-tracking
We have been testing different solutions for this problem, and for now Nightwatch's LLM tracking stands our particularly.
They allow you to configure the phrases manually, and have more control to set it up.
You might want to check for yourself: https://nightwatch.io/ai-tracking
Thanks for the feedback - we will soon allow users to input their own topics/questions - so I’m not surprised to read your preference for it at all. That being said - one of the challenges is how do you ask questions in the voice of the target persona. That’s the tricky part & why we started with the suggested questions first.
How exactly do you determine which questions you ask and the voice of the target persona?
Do you have some actual data to back-up the approach or you are using hypothetical questions?
Do you have some actual data to back-up the approach or you are using hypothetical questions?
We’re playing with a combination of things (frontier models, open weight/open source models we’re maintaining, data) but those two items are key focus area for development.
Very cool. Thanks for sharing. Are you using APIs or scraping with headless browsers for RAG citations?
APIs so that we're not violating TOS (scraping at scale will get you blocked)
You say this, “ Want to appear more often in AI responses?
Gumshoe offers tactical suggestions to improve your brand’s rank in AI search.” But how? I’m not asking for trade secrets, but is it just about publishing seo-style content on Rover’s website that is optimized for LLMs instead of Google algo?
And what is the moat? Won’t every SEO agency just morph to selling this?
And what is the moat? Won’t every SEO agency just morph to selling this?
Yep pointing out features of someone’s site that aren’t optimized for LLM’s is part of it. But the bigger piece is figuring out what 3rd party content is being used as sources for a relevant topic.
SEO agencies use software like this to help their clients. We are already working with some of them. Someone has to do the work of upgrading one’s website & content & it’s usually an agency. We won’t be offering the services - we’ll partner with agencies to do that.
SEO agencies use software like this to help their clients. We are already working with some of them. Someone has to do the work of upgrading one’s website & content & it’s usually an agency. We won’t be offering the services - we’ll partner with agencies to do that.
so is the idea to slowly fix a situation like Rover's competitors being listed in not just ChatGPT but all the major AI chats within 90 days or so? Are the models really being updated often enough to make these changes happen?
I feel like I might seen those same Rover results for the next 2 years no matter what you try?
I feel like I might seen those same Rover results for the next 2 years no matter what you try?
With the move to RAG and models being open to the Internet now - results are changing more quickly. Also we ran an analysis, and for ChatGPT 4o, 51% of the sources/citations were posted in the last 90 days and 82% from the last 12 months.
I see. I just did a claude search:
https://i.imgur.com/hjHvClK.png
So this is all going to a google ad words bidding type system I'm guessing. There will be the new SEO side but also no one can resist a pay for ranking higher system.
With ads: https://i.imgur.com/fk1Kkct.png
https://i.imgur.com/hjHvClK.png
So this is all going to a google ad words bidding type system I'm guessing. There will be the new SEO side but also no one can resist a pay for ranking higher system.
With ads: https://i.imgur.com/fk1Kkct.png
We fully expect a paid ad model to appear - perplexity seems to be the first in that direction. That being said - we see this more about managing your reputation ala credit score per the other comment than the absolute ranking game that traditional search has become.
This is cool because it kind of gets at the core of transparency about AI models.
I would imagine if this would built a certain way it could not only tell a brand what I models are saying, but could be used to look for bias in models that do credit scores, for example.
P.S. I also used to work at Rover!
I would imagine if this would built a certain way it could not only tell a brand what I models are saying, but could be used to look for bias in models that do credit scores, for example.
P.S. I also used to work at Rover!
I agree, it kind of reminds me of this paper that shows LLMs, just like humans, will preferentially remember information, which can lead to biased outputs.
https://www.pnas.org/doi/abs/10.1073/pnas.2313790120
https://www.pnas.org/doi/abs/10.1073/pnas.2313790120
This is really cool actually. I like how it uses the personas to query the LLMs from different angles. And 250 “queries” for free? That’s amazing value and impressive parallelism . Nice job guys!
This looks very good and fast, congrats! One question; how are you going to achieve to appear more often in AI responses?
If you click on the AI Recommendations button when signed in (click signin in the upper right to create an account), then you’ll see the recommendations on how to improve. Hope that helps!
I love this approach. It feels far more like brand and messaging tuning and less like search engine algorithm hacking.
why would someone want this data?
Seems this could be very helpful for pharma, like for market segmentation
Great stuff. Could be very useful for market segmentation in Pharma
We're Todd and Patrick, the founders of Gumshoe (https://app.gumshoe.ai/go). Between us, we have like 50 years of experience in early-stage startups. For better or worse, I helped build one of the original meme sites (ICanHasCheezburger), so it’s partially my fault that there are so many cat photos on the internet. Around the same time, Patrick built Starwave (which is now ESPN Fantasy Sports) and later cofounded UrbanSpoon.
We’re now building Gumshoe to help companies understand how AI talks about their brand.
As AI search tools like ChatGPT and Perplexity have become more common, they’ve changed digital marketing. SEO used to be the primary focus, which led to the hyper-optimized kind of gobbledygook you see on the internet today. Now, marketers need to think about the broader context in which their brand is discussed. This shift is an opportunity for the internet to get better: results could be less "optimized", more nuanced, and ultimately more useful. At the same time, it also introduces challenges. Marketers want to know what AI is saying about their brand and how they can influence it. We want to help marketers share their products through AI, without it feeling forced. Ultimately, everyone wins if the LLMs recommend the best product for you.
The idea for Gumshoe came from a conversation with a friend who founded a large consumer app (Rover, the dog-walking platform). They’ve spent years working on SEO, but when we asked ChatGPT about finding a dog walker, it listed his competitors in the same sentence as Rover. That set him off on a deep dive, trying different prompts to figure out when he was winning and when he wasn't.
Gumshoe automates that process. We run hundreds of conversations with popular LLMs on behalf of our users. Given a brand and a list of relevant topics, we generate search personas, create questions they might ask, and analyze how different AI models respond. The result is a representative sample of what LLMs say about that brand.
What’s different about our approach? Traditional SEO is focused on individual pages, but AI search is more context-driven. While LLMs are trained on fixed data, many RAG implementations seem to prioritize high-quality, concise, and objective content. We’re still researching how LLMs weigh information and we’d love to hear from the HN community about your insights and experiences with AI-driven search. Ultimately, our goal is to make the internet better in the future, so would love your thoughts on how to make sure the best results get surfaced organically in AI search tools.
If you’re curious, we'd love for you to check out Gumshoe and share your feedback. We're here to answer any questions and eager to learn from your perspectives!
Cheers, Todd (sawickipedia) & Patrick (patricko)