In partnership with
You already have a take on which AI lab ships next.
Claude or Gemini? OpenAI or Anthropic? GPT-7 before year-end or not? If you read tech newsletters, you've already formed opinions on all of it.
Kalshi has real-money markets on which AI model leads benchmarks this week, which lab ships AGI first, when Anthropic releases Mythos, whether OpenAI raises ChatGPT pricing, and which company has the best coding model at year-end. These aren't abstract questions — they're live markets with real money on both sides, moving as labs ship, benchmarks drop, and announcements land.
The edge belongs to whoever actually follows this space. Not the casual observer — the person who reads model cards, tracks evals, and notices when a new release outperforms the field before the mainstream press catches up.
That person has a genuine edge. If that's you, Kalshi lets you act on it.
🗣️ This week’s stories
Google's News AI pilot offers publishers promotion in AI-powered overviews within Google News and Gemini, but the catch is they have to grant Google broad rights to use their content, including potentially for training its AI models.
Google is running a World Cup brand campaign featuring Lamine Yamal, Tim Howard and Zlatan Ibrahimovic across TV, YouTube and paid social, aimed at convincing users they can ask Google Search anything now that Gemini powers AI Mode and Overviews.
OpenAI made its debut at the Cannes Lions advertising festival, pitching its young ChatGPT ads business and its Codex coding tool to marketers as it tries to build a new revenue stream ahead of a planned IPO. Ads currently only show to free and $8-a-month Go users, and OpenAI has reportedly told investors it could grow advertising into a $100bn business by 2030, a target some ad executives are sceptical it can hit in just four years.
Alongside OpenAI pitching ChatGPT ads, a wave of AI visibility startups turned up to sell solutions, including Profound, which tracks how AI answers describe and cite brands and made its Cannes debut after hitting a $1bn valuation, and Adobe, which bought Semrush for $1.9bn this year and was promoting a tool for making brands more visible to AI models.
🏃 I mapped who AI cites about Strava and the running category
The sources AI pulls from when it answers running app questions.
In recent weeks, we have launched many exciting features within Obsero, and one of them is the export feature.
We see a significant amount of customers who want to access their data outside of the platform and run analysis via LLMs. We had a fast follow-up of our API launch and next week we're due to launch our MCP, which is extremely exciting.
Strava is a dominant player in its category and has amassed a huge amount of coverage online to help train AI models. It also forms part of the retrieval layer when AI systems have to look for more recent information to help synthesise their answers.
I wanted to know who actually gets cited when AI answers questions about running apps. Not who gets mentioned. Who gets cited. The pages and domains the models pull from when they build an answer.
So I took the raw citation data, pointed an AI agent at it and built a dashboard. No code written by me. Here is the full process that took me 20 minutes to build. I’ve also included the HTML file as a downloadable link at the end.
Step 1: Capture your prompt data using Obsero
The starting point. Strava leads share of voice at 40%, but what is driving this?
Strava sits top of the category for share of voice at 40%, ahead of Nike Run Club, Runna and Garmin. Strong position. But share of voice tells you who gets named, not which sources the models trust to get there. That second question is the one worth answering.
Step 2: Export the raw citations
The data exporter was launched a couple of weeks ago in Obsero and is proving to be a popular feature.
Every AI answer cites sources. I wanted all of them. I picked the Strava organisation and the citations export. That gives one row per citation: the URL, the domain, the model, the topic and the date.
Step 3: Run the export
JSON, citations, go. The export builds in the background.
I exported as JSON so a big file kept its structure, then kicked it off. It runs in the background while you carry on.
Step 4: Download the file
188,236 citations, ready to pull down.
188,236 citations. Three months of data across seven different AI models. Downloaded as a two part file.
Step 5: Point the AI at the folder
Cowork, pointed straight at the data folder where I have saved the exports.
Then I opened Cowork, pointed it at the folder holding the export and told it what I wanted. From here it is a conversation, not a build ticket.
Step 6: Build the look first
A design system first, so the rest looks like Strava.
I started with a design system. Brand colours, the right type, sensible spacing. Do this first and everything you build after looks like Strava, not a default template. You'll most likely save this as a skill with your brand colours, so you do not have to keep running it manually each time.
Step 7: The initial build
First build. The brief, turned into a working report.
Then the real ask. A citation trends report. Two tabs for domain and page level, filters for model, topic and date, a chart of the top movers and a table of the top 200.
Step 8: Page level, the individual URLs
Page level. The exact URLs the models pull from.
Page level shows the exact URLs getting cited. The big best running apps round ups do most of the heavy lifting. Tom's Guide, Marie Claire, the Independent.
Step 9: Domain level, the bigger picture
Domain level. Reddit is the biggest source in the category by a distance.
Roll it up to domains and the pattern is clear. Reddit is the single biggest source at 7.4%, then YouTube, then the app stores. The models lean on community and video far more than any brand's own site.
Step 10: Does the page even mention Strava?
The honest column. Half the top pages never mention Strava.
This is the column that matters. For every cited page, does it actually mention Strava, yes or no. Roughly half the top pages do not. That is the gap.
Step 11: What is rising and what is fading
Declining pages on one side, emerging and brand new on the other
Then a trends view. Which pages are climbing, which are sliding, which are brand new. The big Tom's Guide round up is on the way down. A handful of new pages are coming from nowhere and picking up citations fast.
Step 12: Turn it into an action list
The outreach list. Influential pages that leave Strava out.
From insight to action. The top five publishers cited heavily in the category that do not mention Strava. The Independent, Men's Fitness, a couple of fast rising blogs. That is your outreach list, ranked.
Step 13: Pull it together in one view
The overview. Every headline number in one place.
Finally, the overview. The headline numbers up top. 188,236 citations, 5.1% from Strava's own pages, 71% of crawled pages mentioning the brand. Owned versus earned, share of voice, citations over time.
Step 14: Refine until it reads at a glance
Refined. 100% stacked bars so the splits read in a single glance.
Then I tightened it. Owned share is small, so I split it out properly, switched the splits to 100% stacked bars and added owned versus earned by model and brand presence by model. The aim is simple: understand it in one look.
Raw export to a branded, filterable dashboard in 20 minutes. You can run this yourself, for free - using Obsero.
Download the dashboard for free below 👇
🦁 AI search was hotter than the Cannes sun
A blue robot hoists the Cannes Lions trophy aloft on a spotlit podium while a crowd cheers - AI took centre stage at this year's festival.
Mike, Obsero’s Chief Customer Officer, was there across the week, sitting in on talks and meeting CMOs (when he wasn’t partying with marketing guru Mark Ritson) who are all asking the same question: what can we do to show up in AI search platforms, and how do we measure success?
Speaking to Mike over the weekend, that was the hot topic at Cannes this year. The way discoverability is changing was described as tectonic, an earthquake nobody expected. The New York Times covered it well, calling the old search contract (users searched → links appeared → traffic flowed) effectively dead.
The Forum Stage at Cannes Lions in full swing.
CMOs in the Spotlight: marketing leaders from Heineken, Hinge and Sephora on stage.
One session from Google was very interesting to Mike. The framing was built around E-E-A-T, experience, expertise, authoritativeness and trust, as the foundation for what gets prioritised now.
Showing up in AI search pulls in technical, content, brand, PR and creator strategy all at once, and the real question is who in your organisation owns that soup to nuts.
Most brands don't have an answer yet.
That shift is exactly what we built Obsero for. Not just a platform, but a partner.
Good to see the rest of the marketing industry starting to discuss a topic that has been on our minds since 2022.
That is it for this week.
Until next time.
Andy