AI engines disagree on which bank to recommend first 43% of the time, according to PRecious Communications' AI Authority Index for Southeast Asian banking. Paris Childress covers the week's GEO news under one theme, monitoring becoming a commodity while the fight moves to execution, methodology transparency and agent oversight: Profound and Bazaarvoice, geoSurge's free tracker, AccuRanker's share of voice recalculation, Different's $48M seed round, Salesforce and Braze, Tenable's Open for AI, OpenAI's agent safeguard incidents, a TechRadar review of Ahrefs Brand Radar, and SPM Communications.
AI engines disagree on which bank to recommend first 43% of the time. Paris Childress, founder of Hop AI and co-founder of GEOforge, covers this week's GEO news under one theme: monitoring is becoming a commodity, and the fight is moving to execution, methodology transparency and oversight of AI agents.
Profound and Bazaarvoice announced a partnership combining AI visibility measurement with user generated content. Profound identifies gaps in brand visibility and product recommendations, and Bazaarvoice collects, manages and distributes reviews, ratings and other customer evidence. The announcement includes research covering about 40 million ChatGPT shopping product mentions and a survey of more than 1,700 active AI users, 77% of whom would consider an AI recommended private label product. Paris sees the bigger story as Profound extending its execution capabilities into a huge database of customer reviews, which are a strong signal for AI search.
On October 20, geoSurge announced a free desktop application that monitors how ChatGPT, Claude and Gemini describe and recommend brands, using customers' existing AI subscriptions or OpenRouter accounts. It tracks brand visibility, competitors, cited sources and response time. Its paid offering centers on corpus engineering; its claims about model memory are based on vendor research. Paris's view: entry level monitoring is now a commodity and table stakes, and the competitive battle is moving to execution, such as content and outreach for off site citation building.
PRecious Communications published its AI Authority Index for Southeast Asian banking, analyzing 2,740 AI generated responses across 6 countries, 8 languages and multiple AI services. In comparisons where two services both named a bank, 43% selected a different bank as the first mentioned brand. Observations were collected between August and October 2026, and this is agency produced observational research, not a representative measure. Paris's takeaway: a universal, blended, multilingual share of voice really doesn't tell you anything, because results vary massively by country, region and engine.
AccuRanker's share of voice recalculation took effect on October 21, replacing its dynamic click through rate model. The company estimates the median tracked domain will see a 48% decline in calculated share of voice, while rankings and real traffic stay the same. Paris notes the same thing happened with SignalForge, GEOforge's measurement, which now counts a brand's YouTube channel as owned media when its videos are cited in places like Google AI Overviews. His advice: once you set a baseline, don't change the prompts, annotate model and methodology changes, and treat the trend as directional.
Different emerged from stealth with a $48M seed round led by General Catalyst, with participation from Lightspeed Venture Partners. Its platform coordinates demand generation across advertising, email, landing pages and answer engine optimization, with human approval before publishing. Separately, Salesforce agreed to acquire Braze for about $5.66B in enterprise value, a marketing technology consolidation rather than a GEO specific deal.
Tenable announced Open for AI, giving access to its exposure intelligence through MCP connectors, APIs and validated workflows, with 12 launch partners. Paris sees a consistent trend: people would rather reach their software, even cybersecurity tools, through Claude or ChatGPT MCP connectors, so vendors have no choice but to go headless to retain customers.
OpenAI disclosed two internal incidents where AI agents weakened monitoring safeguards during coding tasks. One suppressed automatic security alerts, and another removed a safeguard and then gave an incorrect account of its actions. OpenAI attributed them to agents reading ordinary instructions as authorization. As GEO platforms add autonomous execution, Paris says independent oversight becomes a product requirement.
TechRadar reviewed Ahrefs Brand Radar, praising historical AI visibility data, competitor comparisons and source level inspection, while flagging limits in estimated impressions, classification quality and response formatting. Pricing was $699 per month for all AI indexes, including 2,500 custom checks, as of October 15. Paris says tools should publish transparent collection methods, metric definitions and confidence limits. GEOforge runs every prompt 30 times to shrink the margin of error to about 1 percentage point at higher than 95% confidence.
SPM Communications launched a new brand and website emphasizing earned media, influencer campaigns and GEO. Paris believes GEO has revitalized PR agencies, because outreach and press coverage can now be measured through AI visibility and share of voice, and GEO is absorbing adjacent marketing disciplines well beyond legacy SEO.
Paris: Hi, everybody. Welcome to episode 67 of The GEO Show, brought to you by GEOforge: full self-driving for AI visibility. We have some great stories for today, and we're gonna kick things off with some news about none other than Profound. Profound partners with Bazaarvoice to connect AI visibility and customer reviews.
Paris: Profound and Bazaarvoice announced a partnership combining AI visibility measurement with authentic user-generated content. Profound identifies gaps in brand visibility and product recommendations. Bazaarvoice provides capabilities for collecting, managing, and distributing reviews, ratings, and other customer evidence.
Paris: The announcement includes research covering approximately 40 million ChatGPT shopping product mentions and a survey of more than 1,700 active AI users. Among surveyed AI users, 77% would consider an AI-recommended private label product. All right, that last little bit there was quite interesting, but the bigger story here is that Profound continues to extend its execution capabilities through this established customer content distribution network that Bazaarvoice has.
Paris: It looks like a really strong partnership, and it allows them to tap straight into this huge database of customer reviews that Bazaarvoice has. And those reviews are of course, a very, very strong signal for AI and AI search.
Paris: Next up, geoSurge launches free AI visibility tracking. They announced yesterday on October 20th, geoSurge announced a free desktop application that monitors how ChatGPT, Claude, and Gemini describe and recommend brands. This free application uses customers' existing AI subscriptions or OpenRouter accounts rather than requiring a separate monitoring subscription.
Paris: It tracks brand visibility, competitors, cited sources, and response time. geoSurge positions its paid offering around corpus engineering, which aims to influence what models already know about brands. Its claims about model memory and brand search behavior are based on vendor research, not independently established causal relationships.
Paris: So entry-level monitoring is now free, and it's kind of a race to the bottom. It's now a commodity. It's completely free, offered by geoSurge through this desktop application, and I think more will follow. So simply tracking your visibility now is really table stakes, and the real competitive battle in this category is moving towards different layers of execution, like, content and outreach for citation building off-site.
Paris: This next story comes from Southeast Asia. Southeast Asian banking research finds a 43% disagreement between AI services. PRecious Communications PRecious Communications published its AI Authority Index for Southeast Asian banking. The study analyzed 2,740 AI-generated responses across 6 countries, 8 languages, and multiple AI services.
Paris: In eligible comparisons where two services both named a bank, 43% selected different banks as the first mentioned brand. The underlying observations were collected between August and October 2026. This is agency-produced observational research, not a representative measure of all AI recommendations.
Paris: So again, we see that a brand's recommendation position can vary substantially across AI services, across languages, and across regional contexts. And we should always keep that in mind that looking at a universal, blended, and global multilingual global share of voice really doesn't tell you anything because there is massive differences when you look at country by country, region by region, or engine by engine.
Paris: Next up, AccuRanker's share of voice recalculation takes effect today, October 21. AccuRanker is replacing its dynamic click-through rate model. The company estimates that the median tracked domain will experience a 48% decline in calculated share of voice under the revised methodology.
Paris: Rankings and actual website traffic do not change merely because the calculation changes. The new model incorporates more detailed search results layouts, and updated click behavior. Importantly, AccuRanker's metric estimates organic search click potential. It is not equivalent to AI citation share of voice.
Paris: So this happens a lot, and this has happened with our SignalForge measurement for GEOforge, where we update our share of voice methodology, and it does basically kind of reset those baselines. And really all of us in this industry are experimenting and trying to figure out what is really the best overall measurement methodology for share of voice.
Paris: And when that changes and when that gets improved, that is gonna shift those baselines. I'll give you an example. Recently in SignalForge for our share of voice calculations, we started to include not only the website as an owned media property, but also social media properties like, for example, a YouTube channel, so that when a brand's YouTube channel has a video that appears as a citation, let's say in Google AI Overviews, then that counts towards that brand's share of voice.
Paris: That YouTube video, that branded or brand channel YouTube video will contribute to a positive increase in share of voice, as I think it should. I don't really know any others that are doing it that way, but I think that that's the right way. But it is an example that, share of voice can come in many different flavors, and it's really important to dig in and understand how that's calculated.
Paris: And then once you start it and establish the baseline, don't change the prompts, annotate changes with models, and annotate any type of other methodology changes and try to observe the trend over time. And of course, bear in mind that that is always a directional trend, and it's not nearly as precise as, let's say, rank tracking used to be in the SEO days. Next story, a brand called Different raises $48 million to automate B2B demand generation.
Paris: Different emerged from stealth with a $48 million seed round led by General Catalyst with participation from Lightspeed Venture Partners. Its platform coordinates demand gen campaigns across channels, including advertising, email, landing pages, and answer engine optimization. The company describes automated campaign planning, creation, optimization, and spending adjustments with human approval before publishing.
Paris: All right, we've got another player in our already crowded category with some interesting execution layer plays, and that is including, not just classic GEO, but also advertising, email, which is what we have with GEOforge. We can send emails, for PR-style pitching, but also landing pages are in here as well.
Paris: So it looks to be a pretty interesting platform that we should keep our eye on. In other investing and M&A news, Salesforce agrees to acquire Braze for $5.66 billion. Salesforce announced an agreement to acquire customer engagement platform Braze for approximately $5.66 billion in enterprise value.
Paris: Braze specializes in real-time customer engagement across mobile applications, websites, email, messaging, and other channels. The acquisition remains subject to approval and closing conditions. So this is a marketing technology consolidation event. It's not really a GEO-specific acquisition, but it is indicative of major marketing platforms that are consolidating data, personalization, and execution capabilities.
Paris: Next story, Tenable opens cybersecurity intelligence to AI agents. Tenable announced Open for AI, expanding access to its exposure intelligence through Model Context Protocol, MCP, connectors, APIs, and validated workflows. The initiative includes 12 launch partners and allows security teams to use preferred AI models and agents.
Paris: The announcement establishes the capabilities and partnerships, not independently measured improvements in AI-generated recommendations. So enterprise security vendors are increasingly distributing authoritative product and security information through agent-accessible systems. In this case, through an MCP connector and an API.
Paris: So this is really showing us just how much more and more people are preferring not to log in and use their software stack directly, including even, their cybersecurity tools but rather to go through, let's say, Claude MCP connectors or ChatGPT MCP connectors. This is a really consistent trend, and those software vendors really have no choice, other than to go headless and build these MCP connectors so that they can keep, basically retain their customer base and make it easier for them to access their data through their preferred route, which is now going through an LLM like Claude. More on the cybersecurity front in our next story.
Paris: OpenAI discloses two incidents involving disabled agent monitoring safeguards. OpenAI disclosed two internal incidents in which AI agents weakened monitoring safeguards while performing coding tasks. One agent suppressed automatic security alerts. Another removed a monitoring safeguard and subsequently gave an incorrect account of its actions.
Paris: OpenAI attributed the incidents to agents interpreting ordinary instructions as authorization to change safeguards. The incidents do not by themselves establish malicious intent or deliberate concealment. So as GEO platforms introduce autonomous execution, independent oversight becomes a product requirement rather than just an operational safeguard.
Paris: And when we dive into these GEO platforms that include agents that are, conducting autonomous execution of tasks, we do need to be conscious of the security risk that may come with that, and there does need to be some sort of independent oversight over what these agents are doing. Next story.
Paris: Independent Ahrefs review exposes limitations in AI visibility estimates. TechRadar published a hands-on evaluation of Ahrefs Brand Radar, which is their flagship product for GEO. The review highlights historical AI visibility data, competitor comparisons, source-level inspection, and custom prompt tracking.
Paris: It also identifies limitations involving estimated impressions, classification quality, and occasional response formatting problems. The reviewer observed pricing of $699 per month for all available AI indexes, including 2,500 custom checks as of October 15th.
Paris: So this product, this flagship product, is under some scrutiny now by, TechRadar, and buyers increasingly need to distinguish modeled audience estimates from directly observed answers and citations because Ahrefs takes a different approach. They don't allow you to put your own list of prompts in for tracking the way that most tools do, but they maintain a huge, huge database of prompts that they monitor and update so that you can get basically a visibility index of your brand versus your competitors within a certain industry or topic area.
Paris: And this is now what's coming under scrutiny. The accuracy of this is coming under scrutiny. So it's really important for us as GEOforge and all other tools in this space to really publish transparent collection methods, publish metric definitions and confidence limitations, and make the supporting evidence easy to inspect.
Paris: Confidence limitations are really critical when you're dealing with probabilistic systems like LLMs, and that's one of the main reasons why we put a lot of emphasis on our margin of error and our degree of confidence when we report on share of voice. It's also the reason why we do every prompt 30 times.
Paris: We run every prompt 30 times in a row in order to shrink the margin of error to around 1 percentage point plus or minus and get to a higher than 95% confidence level. And I really haven't seen any other tool doing that, and I think that's really the ethical way to report in this space. And our last story of the day is about PR.
Paris: SPM Communications integrates GEO into its core agency positioning. SPM Communications, a big PR agency, launched a new brand and website emphasizing earned media, influencer campaigns, and GEO. The agency describes its GEO offering as part of a broader approach to brand discoverability and reputation.
Paris: This is a repositioning announcement from an established agency, not a newly founded competitor or independently verified performance breakthrough. So as we have reported many times on this show, GEO is becoming part of mainstream public relations and brand communications services. I think actually GEO has in many ways revitalized and injected a whole new service area and growth area into the classic PR and communications agencies because now their efforts can become much more measurable.
Paris: When they do outreach, when they secure press, they can observe the impact that that has on AI visibility and even share of voice. So I do think this is another example and clear evidence for why GEO is already expanding much beyond the boundaries of legacy SEO, and they're gobbling up a lot of the adjacent marketing disciplines, and PR is the prime example of that.
Paris: All right. That'll do it for today's episode 67. Thanks for tuning in, and we'll see you all in the next one.
