Last week's issue looked at launches that put a spend meter and an approval default into the product. This week the approval step showed up again, and new data landed on what AI assistants cite. Muck Rack says more than 84% of the AI citations it studied come from earned and third-party sources (GlobeNewswire). Sprinklr opened a beta of six marketing agents, one of which flags exceptions for human review (Business Wire). Oracle says 25 agentic applications are now available, several of which prepare actions for review and approval (PR Newswire).
Also this week, Optimove says its new feature queries a brand's own Snowflake warehouse and gets back only customer IDs (GlobeNewswire). Fibr AI reports citation gains on its own site (PR Newswire), and Contentstack named award winners and published their self-reported results (GlobeNewswire).
Every item below carries a source link. Where a figure or capability comes from a vendor describing its own product, customer, or research, it is labeled a claim and attributed.
The big picture
Three shifts to take into your next planning meeting
- Being mentioned by others still shapes what AI assistants say about you. Muck Rack says paid and advertorial content is 0.3% of the citations it studied, and owned media 15.2% (GlobeNewswire). The sample was consumer prompts, not B2B questions, and Muck Rack sells PR and generative engine optimization tools.
- Review steps keep appearing in agent launches. Sprinklr says its Compliance Agent can flag exceptions for human review before activation (Business Wire). Oracle says several applications prepare actions for review and approval before they run (PR Newswire). Neither release gives pricing or measured results.
- Keeping data in place is becoming a design choice. Optimove says it runs a query against the brand's warehouse and receives back only the customer IDs that meet the criteria (GlobeNewswire). It is an open beta, and the speed figure is the vendor's own.
What AI assistants read
Three items on how brands get cited, and how to read the numbers
AI search research
Muck Rack says earned and third-party sources supply more than 84% of AI citations
Vendor researchMuck Rack's Generative Pulse unit says the October edition of its "What Is AI Reading?" report draws on data from January through September 2026, covering more than 40 million AI responses and 207 million citations, plus a September set of 1,020 unbranded consumer prompts run across ChatGPT, Claude, Gemini, and Perplexity. It says more than 84% of AI citations come from earned and third-party sources, a share that has ranged from 82% to 89% across four editions. Paid and advertorial content is 0.3% of citations, and owned media rose from 13.7% to 15.2%.
The release also says journalism fell from 27% of citations in May to 22%, and that only 212 domains were cited by all four providers. When the same prompt was run twice in one day, ChatGPT repeated 41% of its cited domains and Gemini 45%, compared with 94% for Perplexity. The release says citation behavior may shift as models are updated or retrained. The sample is consumer prompts, the 84% depends on Muck Rack's own source classification, and we did not open the full methodology.
AI search tooling
Fibr AI launches Agent Experience Optimization and reports citation gains on its own site
Vendor claimFibr AI says it launched Agent Experience Optimization (AXO), described in the release headline as turning brand websites into engines for AI visibility and agent action. The company reports that average daily AI citations rose from 41 in its first launch week to 114 during October 1 to 7, 2026, after it made 102 pages on its own site agent-ready, and that 64 of those pages had been cited at least once.
The release gives no pricing or availability. It does not define a citation, name the platforms counted, or describe a control group, and it links to a six-week test methodology that we did not open. CEO Ankur (AJ) Goyal is quoted: "AXO helps brands appear more often in AI answers, with greater control over the information behind them through their own websites."
Case studies
Contentstack names its award winners and publishes their self-reported results
Vendor claimContentstack says Philips rebuilt 1.9 million pages across 90 locales and 45 languages as structured content in 18 months, consolidated more than 75 page types into 35 templates, and saw conversions rise approximately 30%. The release gives no measurement period or method for the conversion figure. It also says a two-site build by the agency Huge for Highstreet Insurance Partners lifted website visits 47% and leads 137% year over year.
Per the release, customers and partners nominate themselves in one of five categories, and a jury of six Contentstack executives picked the winners. The page does not say any metric was audited. CEO Neha Sampat is quoted: "Every marketing team is trying to work out what agentic means for them."
Agents with a review step
Two vendors describe approval and exception handling in their agents
Marketing agents
Sprinklr opens a beta of six marketing agents, including one that flags exceptions for human review
Vendor claimSprinklr says the beta covers six agents: Ideation, Creative, Campaign Execution, Compliance, Community, and Budget Optimization. Per the release, the Compliance Agent checks content against brand standards, organizational policies, and regulatory and channel requirements, and can flag exceptions for human review before activation. The Campaign Execution Agent can connect with compliance checks before launch. The release says people stay in control of strategy, creativity, approvals, and brand decisions.
Customers who want to join the beta contact their account team. The release gives no general availability date, pricing, or customer results.
Enterprise applications
Oracle says 25 agentic applications across Fusion ERP, HCM, SCM, and CX are now available
Vendor claimOracle says twenty-five Claw-powered Agentic Applications are now available, and the release names eight of them, including Ledger, Workforce Orchestration, Service Health Check, and CPQ Test-to-Production Setup & Configuration Migration. Per the release, Ledger actions are "approved and authorized, within organizational policies and controls," and Workforce Orchestration actions "are approved through organizational review." Oracle says its AI Agent Studio includes "built-in observability, ROI measurement, and safety controls."
The release gives no pricing, customers, or measured results, although its subhead promises measurable business outcomes. Chris Leone, executive vice president of Applications Development, is quoted: "The next era of enterprise AI is all about delivering software that increasingly does the work."
Plumbing worth knowing about
Two notes on where data and permissions sit
- Optimove says its Zero Copy Customer Lists builds audiences without customer data leaving the brand's Snowflake environment. Per the release, Optimove runs a SQL query against the brand's warehouse and receives back only the customer IDs that meet the criteria. It is in open beta, rolling out incrementally, with no pricing or general availability date. Optimove also says refreshes that once took three days now finish in 40 minutes, which is its own figure. CEO Pini Yakuel is quoted: "That's a governance gap, and the answer was never going to be another pipeline." (GlobeNewswire, October 26)
- Oracle says Fusion Companion works through supported AI tools such as ChatGPT while respecting established security, permissions, and controls. The release gives no availability date for it (PR Newswire, October 26).
The week ahead
Rollouts with no published dates
- Sprinklr's Marketing AI Agents beta is open now through account teams. No general availability date or pricing is given (Business Wire).
- Optimove's Zero Copy Customer Lists is rolling out incrementally in open beta. No date for general availability is given (GlobeNewswire).
- Fibr AI has not published pricing or availability for AXO. The release says only that it launched (PR Newswire).
Put it to work
What to do with this, this week
- Name the approver for each agent action before you test it. Sprinklr and Oracle both describe review steps in their agents (Business Wire, PR Newswire).
- Run your buyers' top questions through ChatGPT, Claude, Gemini, and Perplexity and write down which sources appear. Muck Rack used those four assistants for its 1,020 prompts (GlobeNewswire).
- Ask any AI visibility vendor to define a citation and to show a comparison group. Fibr's numbers come from its own site with no stated control (PR Newswire).
- Ask for the baseline and period behind any case study percentage. The Philips conversion figure comes with neither (GlobeNewswire).
- If you build audiences from warehouse data, ask whether the tool needs its own copy of that data. Optimove describes a query-only design (GlobeNewswire).
Every claim above carries a source link. Figures attributed to vendors, to companies describing their own customers or results, or to research a vendor ran or published are their claims, not independently verified facts, and are labeled as such. We did not open Muck Rack's full methodology or Fibr AI's six-week test page, and the Sprinklr release page shows no date. Details of beta products may change. Coverage window: October 26 to November 2, 2026. Compiled November 2, 2026.