# REX Radar — 2026-11-16
Persona: REX, Footprint SEO SME (Matt Diamante / Get-Found school). Ingest digest: 2026-11-16T14:30Z (48h window — 8 new items: Search Engine Roundtable 4, Google Search Central 1, blog.google Search 1, Search Engine Journal 1, Growth Memo 1) + search-ingest sweep. **Sweep note:** primary `web_search` provider returned 429 quota errors again this run; the ingest='search' sweep completed via Google News RSS over all 14 search sources (Search Engine Land, Barry Schwartz, Danny Sullivan, John Mueller, Search Off the Record, Aleyda Solis/SEOFOMO, Crawling Mondays, Edward Sturm, Matt Diamante, Nathan Gotch, Whitespark/Darren Shaw, Exploding Topics). Last radar: Nov 5 (no radar Nov 6–15). **Module 99 published today** — claiming and running your business profile inside ChatGPT.

## Status Board

| # | Signal | Tag | Source | Date |
|---|--------|-----|--------|------|
| 1 | **ChatGPT Business Hub is live — claim your business, read your ChatGPT clicks** — OpenAI gives businesses a claimable profile inside ChatGPT with mention counts, website-click counts and discovery topics | **HIGH** — Module 99 basis | Search Engine Roundtable (Barry Schwartz) + OpenAI | Nov 10 |
| 2 | **ChatGPT Business Agents roll out — the agent builder pulls from your website** — US ecommerce/online-service businesses get a "Start chat" agent on their profile; verification checkmark confirms owner control | **HIGH** — new surface | Search Engine Roundtable + Lily Ray/Juozas on X | Nov 16 |
| 3 | **Google November 2026 core update is mid-rollout and the volatility landed Saturday** — announced Nov 10, up to 2 weeks; YMYL chatter is reporting informational pages displacing commercial ones | **HIGH** — live update | Search Engine Roundtable (flux post) + Search Engine Land | Nov 10–16 |
| 4 | **Search Analytics API adds branded-query, web-multimodal and multi-country/device filters** — the three Search Console UI filters now available for automated reporting | **MED** — measurement | Google Search Central (Daniel Waisberg) + SER | Nov 16 |
| 5 | **Growth Memo Q4 AI-search meta-analysis: 85% of retrieved pages are never cited** — and 61.7% of domain appearances are "ghost citations" with no brand mention | **MED** — doctrine | Growth Memo (Kevin Indig) | Nov 16 |
| 6 | **ChatGPT Ads early data: 1/3 the revenue per visit of paid search** — $0.89 vs $2.45 revenue/session, 0.37% vs 1.17% conversion, as OpenAI ships cost-target bidding | **MED** — budget | Search Engine Journal (Matt G. Southern) | Nov 16 |
| 7 | **Google tests dropping AI Mode from the search bar and the Ask Anything box** — plus AI Mode drops the Gemini model label and pushes Gemini 4 Carbon to Pro | **MED** — surface churn | Search Engine Roundtable | Nov 13–16 |
| 8 | **AI Mode ads expand into follow-up questions** — second-source report; not yet confirmed by Google | **LOW** — unconfirmed | MarkHub24 (via Google News) | Nov 16 |
| 9 | **ChatGPT tests web-only results and a "See more" answer button; Bing tests a Web Mode tab** — the AI answers are quietly re-acquiring classic result formats | **LOW** — format watch | Search Engine Roundtable | Nov 16 |
| 10 | **Housekeeping cluster:** Search Console generative-AI report had a data bug Nov 9–11 (SER, Nov 12); AI Mode temporary threads spotted (SER, Nov 16); Google Ads clarified URL exclusions in AI Max (SEL, Nov 16); Gemini 4 Hydrogen in Search (Nov 14); Bing/Microsoft ad outage notes (SER, Nov 13) | **LOW** | Multiple | Nov 12–16 |

## Signal Deep-Dives

### 1. ChatGPT Business Hub: claim the profile, get click data (HIGH — Module 99 basis)
OpenAI launched **Businesses in ChatGPT** (beta) on Nov 10; Search Engine Roundtable covered it the same day with screenshots. Businesses can **claim their business inside ChatGPT**, manage the information shown, and see **analytics: mention counts, website-click counts, and "discovery topics"** — the recurring conversation themes where their site gets cited. Verification runs by **text, phone call to the main business number, or DNS**. Barry's own read of the click metric (from the help tooltip): *"Number of recorded clicks on links to your website in ChatGPT. Clicks don't confirm website visits or conversions."*

**REX Take:** this is the first time an AI platform has handed us a first-party feedback loop. Until now, AI visibility was modeled or inferred — Module 64 (AI visibility measurement) and Module 95 (the 0.62% reality check) both had to lean on sampling and proxies. A claimable profile with mention and click counts changes the operating model in three ways: (1) **accuracy control** — the business details ChatGPT quotes become something we can correct instead of watching go stale; (2) **a real denominator** — mentions vs clicks tells us whether we have a retrieval problem (not mentioned) or a brand problem (mentioned, nobody clicks); (3) **defensive necessity** — an unclaimed profile is a profile someone else can describe for you. The claim flow is 10 minutes. Treat it like GBP in 2015: whoever claims first owns the surface.

**Local Watch:** every managed domain is a US business and eligible for the base claim. Start with the properties where an AI answer is most likely to be the first touch — xactinsure.net, floridahealthinsuranceweb.com, pattonroofsystems.com, kennesawproroofing.com, acrepairdaily.com, securityservicesroseville.com.

### 2. Business Agents: the builder reads your own site (HIGH — new surface)
Search Engine Roundtable (Nov 16, 07:31): the **ChatGPT Agent builder** is rolling out gradually to businesses — "starting with eligible ecommerce and online service businesses in the U.S." A profile with an agent gets a **"Start chat"** button; chats stay private from the business ("Business can't see your chats or your personal info — just overall usage trends"). Barry's first look at the Chewy agent showed a blue verification check reading *"OpenAI verified that this profile is managed by an owner or authorized representative"* — and the agent failed a live product query with "I don't have a live Chewy product catalog connected." Lily Ray's note on the builder: *"it pulls a lot of recommendations out of your website so it's pretty low-effort to get it set up."*

**REX Take:** two operational facts, both ours to use. First, **the agent is only as good as what it can reach** — the Chewy failure is the tell. An agent with no product/service feed, no pricing, no service-area page, no hours, no clear "what we do" answers will deflect. That's the same action-ready doctrine as Module 90 (feed-first, agent-ready content) and Module 99's predecessor work on AI Mode booking links — an agent is just a conversational front end onto your structured data. Second, **the verification checkmark is an ownership signal** — it's the AI-surface equivalent of a verified GBP, and it is the argument for claiming now rather than watching a squatting profile collect the traffic. Businesses that write clear service descriptions, pricing ranges, service areas, and FAQ answers in plain HTML will get a useful agent for free. Businesses that don't will get a polite refusal bot.

**Local Watch:** roofing/HVAC/security/restoration are all "online service businesses" in OpenAI's framing but the rollout is explicitly gradual — check eligibility per property rather than promising it.

### 3. November 2026 core update: the flux landed Saturday (HIGH — live update)
Google's **November 2026 core update** started rolling out **Tuesday, Nov 10 around 1pm ET**; Google says it may take **up to two weeks**. Search Engine Roundtable's Nov 16 flux post: chatter was muted at first, then **picked up Saturday, Nov 14** — "the core update may have kicked in 3-4 days after its announcement" — with third-party volatility tools struggling to measure it (their tracking was already degraded) and Search Console performance reporting running laggy. Google's framing: "a regular update designed to better surface relevant, satisfying content for searchers from all types of sites."

The YMYL chatter in that thread is the part worth reading twice: multiple site owners report **informational pages displacing commercial pages** in YMYL SERPs — "the top 20 is pure info content, no commercial... even if the commercial page is superior in quality, UX, info and helpfulness."

**REX Take:** do not touch anything for the first 72 hours of a core update — Module 96's fix-order doctrine and Module 63's core-update playbook both say the same thing, and the reason is now measurable: the tools can't see the update clearly, so any "fix" you make is a guess against a moving baseline. Concretely: freeze content pruning and template changes on the roster until roll-out completes (~Nov 24), snapshot GSC before/after per property, and log changes rather than ship them. The YMYL report is the interesting one — if informational answers are outranking commercial pages, the response for xactinsure.net and floridahealthinsuranceweb.com is **not** to make pages more salesy; it's to make the informational layer answer the question completely *and* carry the next step (quote path, plan comparison, licensed-agent proof). That's Module 92's trust-gap playbook, and it's the right hedge either way.

**Local Watch:** all 12 domains — read-only monitoring this week. No pruning, no redirect sweeps, no template changes.

### 4. Search Analytics API: branded, multimodal, multi-country/device (MED — measurement)
Google Search Central (Nov 16, Daniel Waisberg / Nati Yosephian / Karin Gadasi): the Search Analytics API now supports the three Performance-report filters that previously existed only in the UI — `brandedClassification` for branded vs non-branded comparison, `type=webMultimodal` for Lens/Circle-to-Search/image-upload performance (also in BigQuery bulk export), and a `filterLists` field for multiple countries or devices in one request.

**REX Take:** this is the plumbing for the KPI pivot in Modules 45 and 87. Branded-vs-non-branded in the API means we can automate the single most useful directional metric we have for AI-era SEO — whether branded demand is growing while click share compresses. Multi-country filtering matters for floridahealthinsuranceweb.com's state structure and for any property serving more than one metro. And `webMultimodal` finally gives visual-search a measurable row; today it's small, but it's the first honest baseline we've ever had for it. Implementation note: run the branded pull now while the core update is mid-flight so you have a pre/post split rather than one blended number.

**Local Watch:** growwithfootprint.com reporting stack first (we own the scripts), then roll the branded split into the standard monthly client report.

### 5. Growth Memo Q4 meta-analysis: retrieval is the bottleneck, not content polish (MED — doctrine)
Kevin Indig's Q4 2026 **AI Search Meta Analysis** (Nov 16, 52 studies cited, 32 discussed) — the load-bearing findings:

- **AirOps, 15,000 prompts / 548,534 retrieved pages:** **85% of retrieved pages were never cited**; among cited pages that ranked top-20, **32.9% appeared only for a follow-up ("fan-out") query**, never the original prompt.
- **Growth Memo × AirOps:** the **first retrieved page had a 58% citation rate vs 14%** further down the list; pages whose headings matched the query more closely were cited more.
- **Ghost citations (Growth Memo × Semrush, 3,981 domain appearances):** **61.7% cited without a brand mention**, 13.2% both, 25.1% mentioned without a citation.
- **AI Overviews on commercial queries:** Semrush's 400-keyword US desktop study — coverage rose from **45.75% (Sep 2025) to 88% (Aug 2026)**.
- **Clicks:** Pew's panel found clicks on a traditional result in **8% of visits with an AI summary vs 15% without**; Semrush/Datos found zero-click *share* fell slightly on keywords that gained Overviews — i.e. the effect is format- and query-dependent, not a single percentage.
- **Agentic shopping is earlier than the hype:** Gartner — 31% would let AI narrow household-supply choices, only 11% would let it decide; Visa's benchmark — models passed 57.2% of attempts, only 34.2% of model/scenario combos passed all three tries.

**REX Take:** this is the most useful doctrinal document of the month, and it collapses to one sentence for the roster: **you cannot optimize your way out of not being retrieved.** 85% waste at the retrieval layer means the win is composition, not copy-editing — be in the candidate set for the *fan-out* queries, not just the head prompt. Practically that's Module 4's query fan-out work plus Module 70's modifier mastery: build the adjacent pages (cost, comparison, "how long does it take", "is it worth it", "who's best in [city]") because a third of citations come from a query you never targeted. And the ghost-citation number is the client-facing one: **61.7% of the time AI uses your page without naming you.** Citation is table stakes; brand mention is the campaign. Stop reporting "cited" as the goal and start counting named mentions.

**Local Watch:** re-run the prompt-set sampling per property and split the results into four columns — retrieved / cited / named / clicked — instead of one "AI visibility" score.

### 6. ChatGPT Ads: a third of the revenue per visit (MED — budget)
Search Engine Journal (Matt G. Southern, Nov 16): OpenAI shipped **cost-target bidding** in ChatGPT Ads this week (ads launched February; OpenAI reported a $1B annualized run rate by Aug 31, and >100,000 advertisers). The measured comparison, pooled across 102 sites: **$2.45 revenue per paid-search session vs $0.89 for ChatGPT ad sessions** (US traffic), and **1.17% vs 0.37% transaction rate per session**. The eye-catching "52¢ back per dollar vs $3.40" figure is *not* measured — it's the measured per-visit metric multiplied by average CPCs from one ChatGPT account and 16 paid-search accounts, so it moves with what you pay per click.

**REX Take:** the honest read is "not yet a Google Ads substitute, and don't quote the 52¢ number as data." For the roster the sequencing stays: own the free surfaces first (Signal 1 — claim the profile, read the click data, fix the description) because that traffic costs nothing and compounds, and treat ChatGPT Ads as a test line item after the organic/agent surface is clean. The methodology point matters as much as the numbers — this is Signal 5's lesson again: separate the measured metric from the modeled one before it reaches a client deck.

### 7. Surface churn: Google tests removing AI Mode, drops the model label (MED)
Search Engine Roundtable: Google is **testing removing AI Mode from the search tab bar and the "Ask anything" follow-up box inside AI Overviews** (Nov 13), following earlier tests removing AI Mode buttons from the homepage. Separately (Nov 16): AI Mode **no longer shows which Gemini model it's using** in the interface header, and **Gemini 4 Carbon** is now available to Pro subscribers (previously Ultra only); Gemini 4 Hydrogen arrived in Search around Nov 14.

**REX Take:** read this as distribution tuning, not retreat. Google is A/B testing how hard to push AI Mode into the classic bar while quietly upgrading the model behind it — which means **answer quality changes without any announcement you can track**. Two consequences: any AI-visibility benchmark you run is model-version-sensitive (log the date, per Module 64), and your content should be written to survive a model swap — clear structure, explicit facts, named entities — not tuned to one model's quirks. The disappearing model label is also a reminder that Google will not give us a stable measurement surface; build on first-party data (GSC, Business Hub clicks) wherever we can.

### 8–10. Watch items (LOW)
- **AI Mode ads in follow-up questions** (MarkHub24, Nov 16, two posts) — single non-primary source, not confirmed by Google. Watch only; if it holds, follow-up inventory changes the AI Mode economics story in Signal 6.
- **ChatGPT tests web-only results + "See more" answer button; Bing tests Web Mode** (SER, Nov 16) — both are the platforms re-acquiring classic result formats. If web-only becomes a shipping ChatGPT tab, the zero-click math in Module 87 shifts back toward clicks. Not yet shippable doctrine.
- **Housekeeping:** Search Console's generative-AI performance report had a **data bug Nov 9–11** (SER, Nov 12) — do not use Nov 9–11 AI-report numbers in any client comparison; AI Mode **temporary threads** spotted (SER, Nov 16) — privacy feature, no SEO action; Google Ads clarified **URL exclusions in AI Max** (SEL, Nov 16) — relevant only if a property runs AI Max; Microsoft 365 outage (Nov 13–15) did **not** directly hit Microsoft Advertising.

## Local Watch
- **Claim audit (Signals 1–2):** verify the ChatGPT Business Hub claim status for all 12 domains; claim via DNS where the domain is ours (fastest, no phone dependency).
- **Core-update freeze (Signal 3):** no pruning, no redirect flattening, no template or schema changes on any managed domain until Google confirms roll-out complete (~Nov 24). Snapshot GSC now.
- **YMYL informational-layer check (Signal 3):** for xactinsure.net and floridahealthinsuranceweb.com, list the queries where an informational page outranks our commercial page; the answer is completing the informational layer, not de-optimizing the money page.
- **Four-column prompt audit (Signal 5):** retrieved / cited / named / clicked per property, replacing the single AI-visibility score.
- **Branded split (Signal 4):** pull branded vs non-branded from the API now, so the core-update window has a clean before/after.
- Carry-forward from Nov 5: Module 99's predecessor booking/quote readiness audit on acrepairdaily.com, pattonroofsystems.com, kennesawproroofing.com, securityservicesroseville.com; Aleyda's 83-check audit on the two YMYL properties; redirect-chain flattening (deferred to post-update per Signal 3); protected-URL registration on money pages.

## Task Queue
1. Claim every managed domain in ChatGPT Business Hub — DNS verification where we control DNS (Signals 1–2).
2. For each claimed profile, screenshot the baseline: mentions, clicks, discovery topics — day one of the first real AI-visibility time series we own (Signal 1).
3. Build the post-claim fix list: business description, service list, service areas, hours, pricing ranges — the fields an agent needs to answer instead of deflect (Signal 2).
4. Freeze all on-site changes on managed domains until the core update completes; snapshot GSC per property today (Signal 3).
5. Add the branded/non-branded API pull to the reporting script and run it across the roster (Signal 4).
6. Re-run prompt sampling with the four-column split; report named mentions separately from citations (Signal 5).
7. Update the client-deck methodology slide: measured vs modeled metrics, and retire the "52¢ per dollar" ChatGPT Ads figure before it gets quoted (Signal 6).

## Per-client action items

| Domain | Action |
|--------|--------|
| xactinsure.net | Claim the ChatGPT profile (DNS) and record baseline mentions/clicks (S1). During the freeze, identify informational queries where info pages outrank commercial ones and plan the completeness fix, not a sales-pitch rewrite (S3). Add branded/non-branded split to reporting (S4). |
| floridahealthinsuranceweb.com | Claim (DNS) + baseline (S1). Multi-state structure: use the new `filterLists` country/device API filter for state-level reporting (S4). Track ghost citations — 61.7% of AI uses go unnamed, so count named mentions per state guide (S5). |
| acrepairdaily.com | Claim + baseline (S1); the seasonal AC guides are prime agent-answer content — write explicit service-area, pricing-range and "how fast can you come" answers (S2). Carry forward the booking/quote readiness audit once the update completes (S3 freeze). |
| growwithfootprint.com | Claim + baseline (S1). We own the reporting stack here: implement the branded/multimodal API pull first, then template it for clients (S4). Freeze on-site changes during the update (S3). |
| buildwithng.com | Claim + baseline (S1). Portfolio copy is already service-descriptive — verify the business description and service list the agent will read match the site (S2). No rebuild work until the update clears (S3). |
| pattonroofsystems.com | Claim + baseline (S1). Roofing is a clean agent use case: ensure storm-damage/inspection/quote pages carry the facts an agent quotes (S2). Hold the proximity-grid and booking audits until post-update (S3). |
| kennesawproroofing.com | Claim + baseline (S1). Add explicit service-area + response-time answers for the agent layer (S2). Review-velocity work continues — that's off-site and safe during the update (S3). |
| securityservicesroseville.com | Claim + baseline (S1). Security is an "online service business" — build the agent answer set (licensing, response time, monitoring options) before the rollout reaches it (S2). |
| edricker.com | Claim + baseline (S1) — personal-brand profile inside ChatGPT; keep the description consistent with the LinkedIn/knowledge-panel entity work from Nov 5 (S2). |
| footprintstudioco.com | Claim + baseline (S1); portfolio-appropriate description and service list for the agent (S2). Light touch only until the update completes (S3). |
| gressofab.com | Ecommerce: claim + baseline now — Business Agents are explicitly "starting with eligible ecommerce" so this property is in the first wave (S1–S2). Verify product/service facts the agent can actually reach (S2). Carry forward the Merchant Center feed check. |
| raybaconstore.com | Ecommerce: claim + baseline (S1–S2); same first-wave eligibility. Retire the unverified ChatGPT Ads "52¢" figure and any modeled AI-traffic stats from decks (S6). |

## Sources
- Search Engine Roundtable — "Businesses in ChatGPT Business Hub Live With Click Data" (Barry Schwartz, Nov 10): https://www.seroundtable.com/chatgpt-business-hub-42505.html
- OpenAI — "Introducing Businesses in ChatGPT" (Nov 10): https://openai.com/index/introducing-businesses-chatgpt/
- Search Engine Roundtable — "OpenAI ChatGPT Business Agent - First Look" (Nov 16): https://www.seroundtable.com/openai-chatgpt-business-agent-42536.html
- Search Engine Roundtable — "Google November 2026 Core Update Ranking Flux - What We're Seeing" (Nov 16): https://www.seroundtable.com/google-november-2026-core-update-ranking-volatility-42543.html
- Search Engine Roundtable — "Google November 2026 Core Update Now Rolling Out" (Nov 10): https://www.seroundtable.com/google-november-2026-core-update-42504.html
- Search Engine Land — "Google November 2026 core update rolling out now" (Nov 10) and "ChatGPT citation drops look like unannounced Google core updates" (Nov 16, Google News RSS verified)
- Google Search Central — "New filtering and reporting capabilities in the Search Analytics API" (Daniel Waisberg et al., Nov 16): https://developers.google.com/search/blog/2026/11/search-analytics-api-update
- Search Engine Roundtable — "Google Search Analytics API Adds Branded Queries, Multimodal & Country/Devices Filters" (Nov 16): https://www.seroundtable.com/google-search-analytics-api-addition-42544.html
- Growth Memo — "AI Search Meta Analysis - Q4 2026" (Kevin Indig, Nov 16): https://www.growth-memo.com/p/ai-search-meta-analysis-q4-2026
- Search Engine Journal — "ChatGPT Ads Vs. Google Search Ads: What Early Data Shows" (Matt G. Southern, Nov 16): https://www.searchenginejournal.com/chatgpt-ads-vs-google-search-ads-early-data/597125/
- Search Engine Roundtable — "Google Tests Dropping AI Mode From Search Bar & Ask Anything" (Nov 13): https://www.seroundtable.com/google-dropping-ai-mode-again-42534.html
- Search Engine Roundtable — "Google AI Mode Drops Gemini Label & Rolls Out Gemini 4 Carbon To Pro Subs" (Nov 16): https://www.seroundtable.com/google-ai-mode-drops-gemini-label-42540.html
- Search Engine Roundtable — "Google AI Mode Start Temporary Thread" (Nov 16): https://www.seroundtable.com/google-ai-mode-temporary-thread-42539.html
- Search Engine Roundtable — "ChatGPT Tests Web Results, See More & Other Search Results Formats" (Nov 16): https://www.seroundtable.com/chatgpt-search-result-tests-42537.html
- Search Engine Roundtable — "Bing Tests Web Mode Tab To Show Only Web Page Results" (Nov 16): https://www.seroundtable.com/bing-web-mode-42535.html
- Search Engine Roundtable — "Google Search Console Generative AI Performance Report Data Bug Nov 9 - 11" (Nov 12)
- Search Engine Land — "Google Ads clarifies how URL exclusions work in AI Max for Search campaigns" (Nov 16, Google News RSS verified)
- MarkHub24 — "Google Expands Ads Into AI Mode Follow-Up Questions" (Nov 16, Google News RSS verified — single secondary source, treated as unconfirmed)
- Sweep note: primary `web_search` provider returned 429 quota errors this run; the ingest='search' sweep completed via Google News RSS across all 14 search sources. No new 48h signal-backed items from Danny Sullivan (X), John Mueller (X / Search Off the Record), Aleyda Solis/SEOFOMO, Crawling Mondays, Edward Sturm, Matt Diamante, Nathan Gotch, or Whitespark/Darren Shaw beyond what is listed above.