Google / Gemini Tracking
Search · AI · Monetisation
Module 01
Executive Summary
Top-level directional signals across all monitored baskets
Updated: —
TOTAL AI USAGE PROXY
Cloudflare Radar
—
loading
Window change
—
Method1st half vs 2nd half of window
Directional AI crawler activity on Cloudflare's network. NOT q/q, NOT y/y, NOT paid-user demand or MAU.
GEMINI API USAGE RANK
OpenRouter
loading…
Best Gemini modelrank in top-30 by tokens
WoW changeweek-over-week token change
OpenRouter API token usage proxy. NOT Google total queries, MAU, search revenue, or market share.
Secondary — Cloudflare Radar internet rank
SEARCH VOLUME / DEMAND
KW Planner
-3.4%
q/q
Q/Q change-3.4%
Y/Y change+8.1%
Commercial keyword basket volume; comparable to Google-reported query growth
SEARCH PRICING
KW Planner
+5.7%
q/q
Q/Q change+5.7%
Y/Y change+12.3%
Blended CPC proxy from high-value keyword bid ranges (Finance, Legal, Travel)
LIVE SIGNAL — MODEL USAGE RANKINGS
Top AI Models by API Token Usage — This Week
Real API consumption data · Source: openrouter.ai/rankings · Token usage proxy — not Google queries or MAU
Live data
OpenRouter
OpenRouter data reflects API token usage on OpenRouter's routing network only.
It does not represent Google's total Gemini usage, search queries, MAU, revenue, or market share.
It is a strong directional proxy for developer/API adoption relative to peers on the same platform.
Window
Applies to live API cards and summary strips. Embeds below are fixed 12W.
LIVE SIGNAL — AI BOT TRAFFIC
AI Bots HTTP Traffic — Rolling Window
Real-world AI crawler & bot traffic share by user-agent · Source: Cloudflare Radar
Live data
Cloudflare Radar
LIVE SIGNAL — AI BOT INTENT
AI Bot Crawl Purposes — 12-Week Time Series
Why AI bots are crawling: training, search-indexing, content scraping & other purposes · Source: Cloudflare Radar
Live data
Cloudflare Radar
What changed this month
Gemini search interest (Google Trends) continuing month-on-month decline; ChatGPT widening lead; Perplexity accelerating fastest among all monitored peers
Commercial keyword basket proxy: Finance & Travel cluster stable-to-firm; Tech/SaaS cluster showing multi-month softening signal (Google Trends + Keyword Planner volume indicators)
Keyword Planner bid ranges firming in Finance & Travel clusters; Tech/SaaS bid ranges declining — structural demand rotation signal
AI Overview appearing in 34 of 120 monitored keywords (28%) this month vs 24 of 120 (20%) last month — 8 pp rise in basket coverage
Vertex AI list pricing unchanged MoM; Bedrock Claude pricing reduced; GCP public interest index stable while Bedrock trends modestly upward
Module 02
Gemini Public Signal Monitor
API token usage (OpenRouter) + public search-interest proxies — directional indices only
Updated: —
PRIMARY SOURCE — API TOKEN USAGE
AI Model Rankings by Token Usage — OpenRouter
Weekly token consumption across all models · openrouter.ai/rankings · Gemini vs GPT-4 vs Claude vs Llama
Live data
OpenRouter
Top 30 Models — This Week
| # | Model | Tokens (week) | WoW % |
|---|---|---|---|
| Loading OpenRouter rankings… | |||
OpenRouter token counts reflect API usage routed through OpenRouter only.
They are a relative developer-adoption proxy — not Google total queries,
search volume, MAU, revenue, or absolute market share.
Gemini token share = Gemini tokens ÷ all top-30 tokens on OpenRouter.
CATCH-UP TRACKER
Gemini vs Peers — Catch-Up Tracker
Is Gemini catching up to the others?
Cloudflare + GT
Public data
| Platform | GT Index (0–100) | Window Change ⓘ | Window Change (vs prior half) | GT Rank | Source |
|---|---|---|---|---|---|
| ChatGPT | 86 (GT Index) | +4.8% (window) | +4.8% (window) | #1 GT rank | GT Fallback |
| Gemini AI | 46 (GT Index) | -2.1% (window) | -2.1% (window) | #2 GT rank | GT Fallback |
| Perplexity | 31 (GT Index) | +7.6% (window) | +7.6% (window) | #3 GT rank | GT Fallback |
| Claude AI | 22 (GT Index) | +4.8% (window) | +4.8% (window) | #4 GT rank | GT Fallback |
| Copilot | — (GT Index) | — (window) | — (window) | — GT rank | GT Fallback |
Data note: GT Index is 0–100 relative scale (not absolute users). Live Google Trends data loaded automatically — falls back to static snapshot if unavailable. Phase 2: paid app-intelligence vendor for absolute DAU/MAU.
Google Trends data is normalized 0–100 and is directional only.
It does not measure sessions, MAU, revenue, or absolute usage.
It is sampled and relative — the peak term in the window is always 100.
Do not use this chart as a primary metric. Label: Fallback Proxy.
Module 03
Search Volume / Demand Tracker
Historical search volume proxy — filed click/volume history vs external proxies · Search revenue = volume × price
Updated: —
MODULE A — HISTORICAL VOLUME PROXY
Historical Search Volume Proxy
Filed click/volume history vs external proxies — Google Ads Search Trends / Insights · Baseline level, q/q %, y/y %
Filed Data
External Proxy
Baseline Level (Latest Q)
~8.5B queries/day
Derived from filed click data
q/q Change
-3.4%
Filed volume proxy (latest Q)
y/y Change
+8.1%
Filed volume proxy (latest Q)
MODULE B — VOLUME BUILD BRIDGE
Volume Build Bridge
Filed click/volume history vs external proxy · Rolling correlation · Search revenue = volume × price
Bridge
Correlation
| Period | Filed Volume (Index) | External Proxy (Index) | Δ (pp) | Rolling Correlation (12Q) | Note |
|---|---|---|---|---|---|
| 2024 Q4 | 100.0 | 102.3 | +2.3 | 0.87 | Proxy slightly ahead |
| 2024 Q3 | 103.5 | 105.1 | +1.6 | 0.89 | Strong correlation |
| 2024 Q2 | 107.2 | 106.8 | -0.4 | 0.91 | Proxy slightly behind |
| 2024 Q1 | 104.8 | 103.9 | -0.9 | 0.88 | Good tracking |
| 2023 Q4 | 108.1 | 107.5 | -0.6 | 0.86 | Within tracking range |
Bridge Note:
Rolling 12Q correlation: 0.87–0.91 indicates strong tracking between filed volume and external proxy. External proxy is suitable for forward-looking volume estimates when combined with pricing signals. Search revenue = volume × price. Volume tracker (this module) + Pricing tracker (Module 04) = revenue build.
Module 04
Search Pricing / Ad Demand
Blended top-of-page CPC proxy · basket volume trend · q/q & y/y · Google Ads Keyword Planner
Updated: —
MODULE C — SEARCH PRICING / AD DEMAND
Search Pricing / Ad Demand
Blended bid/CPC proxy by basket · volume trend · q/q & y/y · source: Google Ads Keyword Planner
KW Planner
Google Ads
| Basket | Geography | Blended Bid / CPC Proxy | Price q/q | Price y/y | Avg Monthly Volume | Volume q/q | Volume y/y | Source |
|---|
Quarterly Price Change by Vertical — Bid/CPC % Δ
Q/q price Δ:
+5%+
+2–5%
±2%
−2–5%
−5%+
| Basket | Avg Monthly Searches | Volume q/q | Volume y/y | Competition | Intent | Source |
|---|---|---|---|---|---|---|
| Financial Services | High (100K–1M) | +2.1% | +8.4% | High | commercial | KW Planner |
| Travel & Hospitality | High (100K–1M) | +3.8% | +14.2% | Med–High | transactional | KW Planner |
| Health & Pharma | Med–High (10K–100K) | −0.4% | +3.1% | High | commercial | KW Planner |
| E-Commerce / Retail | High (100K–1M) | −1.9% | −3.2% | Medium | transactional | KW Planner |
| Tech / SaaS | High (100K–1M) | −3.4% | −8.7% | High | commercial | KW Planner + Search Trends |
| Legal Services | Low–Med (1K–10K) | −0.8% | +1.4% | Very High | commercial | KW Planner |
| Auto / Insurance | High (100K–1M) | +1.6% | +6.8% | High | transactional | KW Planner |
CTR / Monetisation — Requires Live Benchmark Campaigns
Not active in Phase 1. CTR and monetisation metrics require live benchmark campaign data, which is not available through public Keyword Planner alone. Activating this tab with fake numbers would mislead the model.
Blended CTR by vertical — requires live benchmark campaign or paid Search Console access
Revenue-per-click (RPC) — not publicly observable; requires enterprise reporting
AI Overview deflection impact on CTR — no public data source covers this in Phase 1
Zero-click rate trends — requires large-scale query log sampling
Phase 2: activate with live benchmark campaigns + approved paid data source
Not shown in Phase 1 to avoid unsupported precision.
Module 05
What It Means for Google Search Revenue
Volume × Price signals · directional only · source: KW Planner + AIO basket audit
Updated: —
Volume signal
q/q +2.1%
y/y +8.4%
KW Planner basket median · directional proxy
Price signal
q/q +3.2%
y/y +7.8%
Blended top-of-page bid proxy · Finance + Travel leading
Net implication
Positive if sustained
Revenue = Volume × Price · both signals directional positive this quarter
Watch item
AIO moving into high-value queries
AIO basket: 34 / 120 keywords (28%) — up from 20% last month · Finance & Legal at risk if crosses 30%
Both volume and price proxies are directional this quarter. The risk that offsets is AIO expansion into Finance and Legal — those verticals account for the highest bid density. Monitor AIO penetration rate monthly against these baskets before drawing a revenue conclusion.
Module 05B
AI Overview Watchlist
High-value verticals only · manual SERP audit · 120-keyword basket
Updated: —
Total tracked queries
120
keywords
Manually curated basket · monthly SERP spot-check · not population-level
High-value penetration
6 / 38
16%
↑ +5 pp MoM
High-value = bid >$8 top-of-page · Finance, Legal, Health, Auto clusters · was 11% last month
By vertical
| Vertical | KWs tracked | AIO present | Rate | MoM | Signal |
|---|---|---|---|---|---|
| Financial Services | 18 | 3 | 17% | ↑ +2 pp | Watch |
| Legal Services | 14 | 2 | 14% | → Stable | Low |
| Travel & Hospitality | 14 | 4 | 29% | ↑ +3 pp | Watch |
| Health & Pharma | 16 | 7 | 44% | ↑ +5 pp | Watch |
| Auto / Insurance | 10 | 2 | 20% | → Stable | Low |
If AIO rises in Finance / Legal / Travel, monitor pricing pressure in those baskets.
LIVE SIGNAL — AI INFERENCE USAGE
Workers AI — Model Usage by HTTP Traffic · 12-Week Time Series
Which AI models are seeing the most real inference demand? Live request volume by model · Source: Cloudflare Radar
Live data
Cloudflare Radar
LIVE SIGNAL — AI INFERENCE BY TASK
Workers AI — Inference by Task Type · 12-Week Time Series
What AI is actually being asked to do: text generation, summarisation, classification, speech & more · Source: Cloudflare Radar
Live data
Cloudflare Radar
Phase 2
Traffic / Volume / Engagement
Requires broader observable coverage or paid traffic / engagement data sourcesTraffic
google.com monthly visit estimates, device split, geographic mix, source-of-traffic breakdown
Monthly visit volume (MoM / YoY) — requires approved web-intelligence panel
Device split (mobile / desktop / tablet) — same paid panel
Geographic traffic mix — same paid panel
Source-of-traffic breakdown (direct / organic / referral) — same paid panel
Volume
Search query volume, impressions, click volume by vertical and intent — not observable from public signals
Exact keyword search volume — Keyword Planner provides tiers only (High/Med/Low), not counts
Impression volume by vertical — requires Search Console or paid panel
Click volume estimates — requires paid data source, significant uncertainty range
KW Planner volume tiers available now in Module 04 (directional only)
Engagement
App DAU/MAU, session depth, retention, YouTube engagement — all require paid vendor coverage
Gemini app DAU/MAU — requires Sensor Tower / data.ai subscription
Session depth and retention — not publicly observable
YouTube watch time, Shorts CPM, CTV engagement — requires paid video-intelligence vendor
ChatGPT / Perplexity app engagement head-to-head — same vendor requirement
Public search-interest proxy available now in Module 02 (GT index, directional only)
Module 08
AI Model Benchmarks
AI model performance benchmarks: image generation quality and agentic capabilities · Source: Artificial Analysis
Live data from artificialanalysis.ai
LIVE BENCHMARK — TEXT-TO-IMAGE QUALITY
Text-to-Image Model Leaderboard
Human preference rankings for text-to-image AI models · Tracks Imagen, DALL-E, Midjourney, Stable Diffusion, and other leading models
Live data
Artificial Analysis
LIVE BENCHMARK — AGENTIC CAPABILITY
AI Agentic Index — Model Performance Across Agentic Tasks
Measures AI models' ability to perform autonomous, multi-step reasoning and task execution · Source: Artificial Analysis
Live data
Artificial Analysis
System
Active Alerts
Signal thresholds crossed since last review cycle
HIGH
—
Gemini Search Interest — Sustained Decline
Google Trends relative interest for "Gemini AI" has shown consecutive monthly declines against ChatGPT baseline. Perplexity is the only peer with accelerating interest. Directional signal only — not absolute engagement data.
WATCH
—
AI Overview Basket Rate — Rising
AI Overview appearing in 28% (34/120 keywords) of our monitored basket, up from 20% (24/120) last month. 8 percentage-point rise in one month. If Finance/Legal verticals breach 30%, monetisation offset thesis is directly challenged.
INFO
—
Finance Keyword Bid Ranges — Firming
Google Keyword Planner top-of-page bid ranges in Financial Services and Travel clusters showing upward direction. Consistent with recovering auction density in high-yield verticals. Supportive of partial revenue-per-query offset thesis.
WATCH
—
AWS Bedrock Claude — List Price Reduction
Bedrock Claude 3.7 Sonnet and 3.5 Haiku list prices reduced vs prior published prices. Vertex AI list pricing unchanged. Bedrock search interest trending modestly upward on Google Trends — competitive pressure signal.
System
Methodology & User Guide
What every tab shows, where the data comes from, and what it does not mean Critical data-source caveats — read before using any number
Cloudflare Radar
Live feed
What it actually is: Cloudflare sits in front of roughly 20% of all websites globally. When AI bots (ChatGPT's crawler, Gemini's crawler, Claude's crawler, etc.) visit those sites, Cloudflare logs the HTTP request. Cloudflare Radar publishes those logs publicly as a time-series chart.
What it does NOT mean: Cloudflare Radar reflects AI bot and crawler traffic popularity — not paid-user demand, not revenue, not how many people are actively using ChatGPT or Gemini. A bot crawling websites to train a model looks identical in this data to a bot powering a user-facing product. Do not use it as a substitute for user engagement data.
Google Trends
Proxy / directional
What it actually is: Google Trends samples a fraction of all Google searches and scores the result on a 0–100 scale. The score 100 means "this was the peak search interest in this time window." A score of 50 means half the peak interest — it does not mean half the search volume of the leader.
What it does NOT mean: Google Trends is sampled and normalised 0–100. It is directional, not absolute volume. You cannot compare a score of 86 for ChatGPT to a score of 46 for Gemini and conclude ChatGPT has nearly twice the users. The numbers tell you the direction (rising/falling) and rough relative order — nothing more. Seasonal events and time-window choice can change the scores significantly.
Google Keyword Planner
Proxy / indicative
What it actually is: Google Keyword Planner is a free tool inside Google Ads. You type in a keyword and it shows you: (a) an estimated average monthly search volume range (e.g., "100K–1M"), (b) a top-of-page bid range (e.g., "$8.50–$14.20"), and (c) a competition level (Low / Medium / High). These figures are pulled from Google's own advertiser-planning data.
What it does NOT mean: Keyword Planner requires a Google Ads account with billing information set up. It provides historical metrics, top-of-page bid ranges, competition levels, and forecasts — but the volume figures are broad tiers (e.g., "1K–10K"), not exact counts. The bid ranges are what advertisers are bidding in the planning tool — the actual cost-per-click in a live auction depends on Quality Score, ad relevance, auction density, and bid strategy, none of which are visible here. These numbers are an indicative planning range, not a real-time price.
Tab-by-tab guide — what each screen shows and how to read it
What this tab is for
This is the first screen you see when you open the dashboard. It gives you a one-page summary of the four most important signals across the whole dashboard so you can decide in under 60 seconds whether anything has changed materially this month.
What each element means
Four KPI cards (top row)
Each card is one headline number with a q/q (quarter-on-quarter) and y/y (year-on-year) change. Clicking any card opens a detail drawer that explains the methodology for that specific number.
- Total AI Usage Proxy (+18.2% q/q): A composite directional score built from Cloudflare bot-traffic trends and Google Trends public interest. It tells you whether total AI-product activity appears to be growing or shrinking. It is not a user count or revenue figure.
- Gemini Share Trend (#2 → #2, −2.1 pts): Gemini's relative position vs ChatGPT, Claude, and Perplexity on the Google Trends 0–100 scale. The number tracks whether Gemini is catching up or falling behind in public interest — not actual market share.
- Search Volume / Demand (−3.4% q/q): A proxy for how Google's core search query volume is moving. Sourced from Google Ads Keyword Planner volume tiers. Directional only — not an exact query count.
- Search Pricing (+5.7% q/q): A blended average of top-of-page bid ranges from Keyword Planner across our highest-value keyword baskets (Finance, Legal, Travel). Rising pricing means advertisers are competing more aggressively for Google search placement. Not an actual CPC.
Colour coding on cards (green / amber / red)
Green = directionally positive for Google's revenue outlook. Amber = neutral or watch signal. Red = directionally negative or declining. These are editorial judgements, not algorithmic scores.
Cloudflare Radar charts (two live embeds)
Two live charts fed directly from Cloudflare's public data explorer. These update daily and cannot be manually refreshed.
- AI Bots HTTP Traffic: Which AI company's crawlers (ChatGPT-bot, Googlebot, ClaudeBot etc.) are generating the most HTTP traffic across Cloudflare's network over the last 12 weeks. Rising lines = that company's bots are crawling more aggressively.
- AI Bot Crawl Purposes: Why those bots are crawling — model training, search indexing, content scraping, or other purposes. This is useful context for the AI arms race but is not a measure of how many users are actively using any of these products.
These charts show bot/crawler traffic only — not paid-user demand or revenue.
"What changed this month" panel
A manually written summary of the most important signal changes since the last monthly refresh. Updated by the research team, not auto-generated. Each line is colour-coded by severity.
Refresh button
Clicking "Refresh Signals" reloads the timestamp and re-renders any calculated values on this tab. The Cloudflare live embeds update on their own schedule and do not depend on this button.
Cloudflare Radar
Google Trends
KW Planner
Manual editorial
What this tab is for
This tab answers one question: is Gemini catching up to ChatGPT, Claude, and Perplexity — or falling further behind? It tracks the relative public profile of each AI assistant using the best public proxy available (Google Trends + Cloudflare bot traffic). Phase 2 will upgrade this to absolute daily/monthly active users from a paid data vendor.
What each element means
Phase 1 banner (blue notice bar)
Tells you upfront that this tab is using proxy data only — not absolute user numbers. Read this before drawing conclusions.
Relative Share / Usage Trend table (primary view)
A four-row table showing ChatGPT, Gemini, Perplexity, and Claude side by side.
- Current Proxy Level (e.g., "86 (GT Index)"): The Google Trends score on a 0–100 scale. 100 = peak interest in the period. ChatGPT at 86 and Gemini at 46 means ChatGPT has nearly twice the search interest — but you cannot say it has twice the users.
- q/q Change: How that interest score has moved quarter-on-quarter. A negative number for Gemini (−2.1%) means fewer people are searching for Gemini this quarter compared to last quarter. Not a user churn figure.
- y/y Change: Same comparison, year-on-year. Perplexity at +210% y/y means its search interest has tripled in a year — a strong directional signal even without absolute user numbers.
- Rank / Share Change: Editorial judgement on competitive positioning. "#2 → #2" means Gemini held its spot but is not catching up. "Accelerating" is a qualitative signal.
Public Search Interest chart (second tab — fallback)
A line chart showing the same four platforms over the last 12 months using Google Trends index values. Marked as "fallback only" because it is normalised and sampled — useful for trend direction but not for precise comparisons. The chart currently uses illustrative directional curves pending live API integration.
What "Phase 2" means here
Phase 2 = we need a paid subscription to Sensor Tower or data.ai (app-intelligence vendors) to get absolute monthly active users, daily active users, and download rankings for the Gemini app vs ChatGPT app. Until that subscription is approved, we use Google Trends as a directional stand-in.
Google Trends (proxy)
Cloudflare bot traffic (proxy)
Phase 2: Sensor Tower / data.ai
What this tab is for
Google's search revenue equation is: Revenue = Volume × Price. This tab tracks the Volume side of that equation. If search query volume is rising, that supports Google's revenue even if pricing is flat. If volume is falling, that creates a headwind. This tab shows how the volume signal is moving using the best publicly available proxies.
What each element means
Phase 1 banner
Confirms that Google Trends is only used as a secondary overlay here, NOT as the main number. The main source is public-filings-derived volume/click history.
Module A — Historical Volume Proxy chart
A line chart showing two data series over time:
- Filed Volume (blue line): An index built from click and query volume disclosures in Google's public 10-K and 10-Q earnings filings. Google discloses "paid clicks" and "cost-per-click" in its quarterly reports — from these you can build a directional volume index. Base = 100 at the start of the visible period.
- External Proxy (green line): Google Ads Search Trends / Insights, which reflects relative search demand by market. Used to sense-check the filed data between earnings quarters.
The three buttons (2Y / 5Y / Since 2006) change the time window of the chart. Longer windows show structural growth trends; shorter windows show recent quarter movements.
Google Trends is NOT used as the main number here. It appears only as a directional overlay.
Three metric cards below the chart (Baseline / q/q / y/y)
- Baseline Level (~8.5B queries/day): An estimate of Google's current daily search query run-rate, derived from filed click data. Treat this as order-of-magnitude context, not a precise figure.
- q/q Change (−3.4%): How the volume proxy has moved this quarter vs last quarter. Negative = volume appears to be declining. Positive = appears to be growing.
- y/y Change (+8.1%): Same comparison over the full year.
Module B — Volume Build Bridge table
This table validates the external proxy. It compares what the filed data says vs what the external proxy said for the same quarter. If they track closely, you can use the external proxy to estimate the next quarter before the filing comes out.
- Filed Volume (Index) / External Proxy (Index): Both rebased to 100 at a common starting point. Values above 100 = higher than the base period.
- Δ (pp): The gap between the two sources in the same period. Small gaps (±2) = strong tracking. Large gaps = the proxy is drifting from reality.
- Rolling Correlation (12Q): A statistical measure (0 to 1) of how closely the two series have moved together over the last 12 quarters. 0.87–0.91 = strong correlation. If this drops below 0.70, the proxy is becoming unreliable.
Public filings (10-K / 10-Q)
Google Ads Search Trends
Google Trends (overlay only)
What this tab is for
This tab tracks the Price side of Google's Revenue = Volume × Price equation. "Price" in this context means the cost-per-click (CPC) that advertisers are paying Google when someone clicks on a paid search ad. Higher CPC = Google earns more per search. This tab uses Keyword Planner bid ranges as a directional proxy for CPC trends.
What each element means
Filters row (Basket / Geography / Competition)
- Basket filter: Lets you narrow the table and charts to one industry vertical (Financial Services, Travel, Health, etc.). The keyword basket is pre-defined — this just filters the display.
- Geography filter: Shows US-only, UK-only, or all geographies. Useful because CPCs differ significantly by market (US Finance CPCs are much higher than UK).
- Competition filter: Shows only High, Medium, or Low competition keywords. High competition = many advertisers bidding = higher CPCs, more reliable proxy for Google's revenue density.
Pricing Trend tab — line chart
Shows how the blended CPC proxy has moved over 8 quarters for 5 major baskets. Rising lines = advertising pricing is firming in that vertical. Falling lines = pricing softening. These are planning-tool bid ranges, not actual auction prices.
Keyword table (9 columns)
- Basket: The industry category this keyword belongs to (e.g., Financial Services, Legal).
- Geography: The market this keyword data covers (US / UK).
- Blended Bid / CPC Proxy: The midpoint of Keyword Planner's top-of-page bid low and high range. E.g., if Planner shows $8.50–$14.20, the blended figure is $11.35. This is the planning-range midpoint, not the actual price paid.
- Price q/q and Price y/y: How that blended bid has changed quarter-on-quarter and year-on-year. Green = firming (good for Google). Red = softening (bad for Google).
- Avg Monthly Volume: Keyword Planner volume tier (e.g., "High (100K–1M)"). Not an exact count — Planner only gives broad brackets.
- Volume q/q and Volume y/y: Direction of that volume tier, quarter-on-quarter and year-on-year.
- Source: Always "KW Planner" or "KW Planner + Search Trends" — confirming the data origin.
Heatmap (Quarterly Price Change by Vertical)
A colour grid showing q/q % change in blended bid for each vertical, each quarter. Red cells = pricing rising (≥+5%). Blue cells = pricing falling (≤−5%). Neutral = within ±2%. This gives you a quick scan of which industries are seeing the most auction-price movement.
Volume Trend tab
Same idea as the Pricing Trend tab but focused on volume tier changes for each basket. Useful as a cross-check: if both price and volume are rising in Finance, that is a strong positive signal for Google's Finance search revenue. If volume is falling while price rises, advertisers are concentrating on fewer, more competitive queries.
CTR / Monetisation tab — Phase 2, locked
Click-through rates and revenue-per-click require live campaign data from Google Ads or paid Search Console access. These are not available from Keyword Planner alone, so this tab is locked until a qualifying data source is approved. No placeholder numbers are shown.
Google Ads Keyword Planner
Phase 2: Live benchmark campaigns
What this tab is for
This is the synthesis tab. It does not add new data — it takes the Volume signal from Module 03 and the Price signal from Module 04, puts them side by side, and tells you whether the combined directional picture is positive or negative for Google's search revenue this quarter. Think of it as the "so what" page.
What each row means
Volume signal row (q/q and y/y chips)
The directional summary of the search volume proxy from Module 03. Green chip = volume appears to be rising. The source note confirms this is the "KW Planner basket median" — a blended average across all tracked verticals, directional only.
Price signal row (q/q and y/y chips)
The directional summary of the blended CPC proxy from Module 04. "Finance + Travel leading" means those two verticals are driving the upward pricing signal — important context because Finance and Travel are the highest-yield clusters in our basket.
Net implication row
Revenue = Volume × Price. If both are positive, the implication is positive. "Positive if sustained" means the signals are supportive this quarter, but one quarter of proxy data is not enough to declare a trend — you need to see it confirmed in the next earnings filing.
Watch item row (amber)
The most important risk offset to the positive signal. "AIO moving into high-value queries" means Google's own AI Overview feature is starting to appear in the same Finance, Legal, and Travel searches that have the highest advertising value. If AI Overview takes the user's attention before they click on a paid ad, Google could earn less per search even if pricing and volume are both rising. This is the core investment risk being monitored.
Analyst note (grey italic text at bottom)
A single-sentence synthesis of the whole tab, written by the research team. Updated each monthly refresh cycle.
KW Planner (Modules 03 & 04)
AIO basket audit (Module 05B)
Manual synthesis
What this tab is for
AI Overview is Google's feature that shows a generated AI answer at the very top of a search results page, above all paid advertisements. If AI Overview appears, users may get their answer without clicking any paid ad — which would hurt Google's revenue per query. This tab tracks how often AI Overview is appearing in the keyword baskets that matter most for Google's ad revenue.
What each element means
Total tracked queries (120 keywords)
We manually check 120 Google search queries every month to see if AI Overview appears. These 120 keywords are hand-selected to cover the verticals most relevant to Google's advertising revenue (Finance, Legal, Health, Travel, Auto). "Manually curated basket · monthly SERP spot-check · not population-level" means this is a sample, not a full survey of all Google queries. The sample is designed to be representative of high-value commercial queries, not all searches.
High-value penetration (6 / 38 · 16% · ↑ +5 pp MoM)
Of the 120 keywords, 38 are classified as "high-value" — meaning their Keyword Planner top-of-page bid exceeds $8. These are the most monetisable queries in our basket.
- 6 / 38: AI Overview was present in 6 of those 38 high-value queries this month.
- 16%: That is 16% penetration of the high-value basket.
- ↑ +5 pp MoM: Last month it was 11%. A 5 percentage-point rise in one month is a meaningful acceleration. If this continues, Finance and Legal queries — the highest-yield clusters — will have significant AI Overview presence within a few months.
By vertical table (Finance / Legal / Travel / Health / Auto)
- KWs tracked: Number of our 120 keywords in that industry.
- AIO present: How many of those keywords showed AI Overview this month.
- Rate: The percentage. Finance at 17% and Travel at 29% are the most important to watch — these are high-value, high-yield verticals.
- MoM: Month-on-month change in percentage points. An upward arrow with "+3 pp" or more is a material acceleration.
- Signal: "Watch" (amber) = worth monitoring closely. "Low" (green) = current penetration level is not yet a material risk.
Interpretation line
"If AIO rises in Finance / Legal / Travel, monitor pricing pressure in those baskets." — This is the core investment thesis test. If AI Overview captures those queries, advertisers may reduce spend in those verticals (because their ads are being displaced), which would lower CPCs. That would be a double headwind: lower volume and lower price in the same verticals.
Cloudflare Radar — Workers AI inference charts (two live embeds)
Two live Cloudflare charts showing which AI models and which task types are being used via Cloudflare's inference service. These supplement the AIO watchlist by showing AI demand trends at the infrastructure level.
These charts reflect inference request share on Cloudflare Workers AI — not total global AI inference volume. They do not measure user counts or revenue.
Manual SERP audit (monthly)
Cloudflare Radar (live)
What this tab is for
This tab is a placeholder for data that we want to track but cannot get from public sources. It lists exactly what is missing and why. No numbers are shown — this is intentional. Showing placeholder numbers would mislead the model.
What will be activated in Phase 2
Traffic (google.com monthly visits)
Requires a paid web-intelligence subscription (e.g., SimilarWeb, Semrush Traffic Analytics). Will add device split, geographic mix, and source-of-traffic breakdown.
Volume (exact search query counts)
Keyword Planner only gives broad volume tiers (High / Medium / Low), not exact counts. Exact counts require Google Search Console access (property-level only) or a paid panel.
Engagement (Gemini app DAU/MAU)
Requires Sensor Tower or data.ai subscription. Without this, we cannot compare Gemini app daily/monthly active users to ChatGPT app in an absolute sense.
All Phase 2 — pending data-source approval
What this tab is for
This tab tracks how Google's AI models (Gemini, Imagen) rank against OpenAI (GPT, DALL-E), Anthropic (Claude), and others on independent third-party benchmarks. Strong benchmark performance is a leading indicator of enterprise adoption and helps assess whether Google is gaining or losing technical ground in AI.
What each element means
Text-to-Image Leaderboard (embedded widget from Artificial Analysis)
An independent live leaderboard published by Artificial Analysis (artificialanalysis.ai), a third-party AI benchmarking company. It ranks image-generation models based on crowdsourced human preference testing: real people compare outputs from different models in blind tests and vote for the better image.
- Higher score = better perceived image quality.
- The widget has four inner tabs: Image Leaderboard (ranking), Image Arena (you can vote), Text to Image Leaderboard (detailed), Image Editing Leaderboard (editing quality).
- This data updates continuously as Artificial Analysis runs new evaluations.
AI Agentic Index (embedded widget from Artificial Analysis)
A benchmark that measures AI models on "agentic" tasks — multi-step reasoning, tool use, and autonomous goal completion. Think of it as: if you give the AI a complex task like "book me a flight, summarise this report, and send a draft email to my client," how well does it perform each step without human help?
- Higher score = better autonomous task execution.
- This matters for enterprise AI adoption. Companies evaluating AI tools for business use care deeply about agentic performance.
- Google's Gemini models compete directly here against OpenAI, Anthropic, and Meta Llama models.
Why these are third-party benchmarks
Using Artificial Analysis data means neither Google nor any competitor controls the benchmark. Independent evaluations are more credible for investment research than company-published benchmarks.
Artificial Analysis (artificialanalysis.ai) — live embeds
What this tab is for
The Alerts tab surfaces the most important signal changes since the last review cycle. Each alert is tagged with a severity level. These are manually curated — they are not auto-generated by an algorithm.
Severity levels explained
HIGH
A signal that has crossed a materially important threshold and warrants immediate review before the next investment decision. Example: Gemini search interest showing sustained consecutive monthly declines.
WATCH
A signal moving in a direction that could become material if it continues. Monitor closely but not yet action-required. Example: AI Overview basket rate rising — not yet at the critical threshold but accelerating.
INFO
A notable data point or context item. Not a warning — just relevant information for the model. Example: Finance keyword bid ranges firming.
Manually curated — updated each monthly refresh
Field glossary — every term used in the dashboard, defined plainly
| Term / Badge | Plain-English meaning | What it does NOT mean |
|---|---|---|
| Proxy / Directional | The number is an approximation from a public tool, not a direct measurement of the thing we care about. It tells you the direction (up / down) but not the precise quantity. | An exact count, an audit-quality figure, or something you can put directly into a financial model without adjustment. |
| Live data / Actual | The data is pulled from a live public feed (e.g., Cloudflare Radar embeds) or is a directly published figure (e.g., vendor pricing pages). It reflects the real current state as published. | Proprietary, licensed, or internally reported data. "Live" here means live public data only. |
| Phase 2 / Unavailable | This metric cannot be sourced from public data. It requires a paid vendor subscription or internal data access that has not yet been approved. | Nothing. No number is shown. Blank = not available, not zero. |
| q/q (quarter-on-quarter) | The change from one three-month period to the next. e.g., Q4 2024 vs Q3 2024. | Month-on-month. q/q compares complete quarters, not the most recent 30-day period. |
| y/y (year-on-year) | The change from the same period last year. e.g., Q4 2024 vs Q4 2023. Removes seasonal effects. | Cumulative growth. A y/y of +8% means this quarter was 8% bigger than the same quarter last year — not that the full year grew 8%. |
| MoM (month-on-month) | Change from last month to this month. Used for the AIO basket because we check it monthly. | Seasonally adjusted. MoM can be noisy — one month is not enough to call a trend. |
| pp (percentage points) | The absolute difference between two percentages. If AIO rate goes from 11% to 16%, that is +5 pp — not +45%. | A percentage change. "5 pp rise" and "45% relative rise" are two different things. We always use pp for rate comparisons. |
| GT Index (0–100) | Google Trends normalised score. 100 = peak interest in the selected time window. Relative to the highest-searched term in the comparison set. | A user count, session count, revenue figure, or market share. It is purely a measure of search query interest, normalised and sampled. |
| Blended CPC / bid proxy | The midpoint of Keyword Planner's top-of-page bid low and high range for a keyword. Calculated as (bid_low + bid_high) ÷ 2. | The actual cost-per-click paid in a live Google auction. Real CPCs depend on Quality Score, auction competition, and bid strategy. |
| Volume tier (e.g., "High (100K–1M)") | Keyword Planner's estimate of average monthly searches for a keyword, expressed as a bracket. "High" means between 100,000 and 1,000,000 searches per month. | An exact search count. Planner does not publish precise figures — only brackets. |
| AIO (AI Overview) | Google's feature that displays a generated AI answer at the top of search results, above paid advertisements. Introduced in 2024. | A paid product, an ad, or anything that Google earns revenue from directly. AIO is a free feature. Its risk is that it displaces paid ad clicks. |
| High-value basket | The subset of our 120 monitored keywords where the top-of-page bid proxy exceeds $8. These are the searches Google earns the most from per click. | The most popular searches. High-value = high advertiser spend, not high search volume. Legal services queries are high-value but not high-volume. |
| Confidence level | What it means practically | How to use it in your model |
|---|---|---|
| Published / Verified | Directly published by the data owner on a public page (e.g., Google's pricing page, vendor pricing pages). You can verify by visiting the URL. | Can be used directly. Note: list prices are not negotiated enterprise rates. |
| Directional | The direction (up/down) is reliable but the magnitude is not. A 5% rise in Google Trends might actually be a 3% or 8% rise in reality. | Use to validate directional thesis (e.g., "pricing is firming"). Do not plug the number into a DCF model as a precise input. |
| Indicative ±range | The number is a planning-range estimate, not a precise figure. Keyword Planner bids are what advertisers enter in a planning tool, not what they actually pay. | Use to assess relative levels across verticals and over time. Build in a ±30% margin of error if using as a model input. |
| Sample-based | The number comes from checking a sample (120 keywords) rather than the full population. The true figure could be higher or lower. | Use for trend direction and relative risk assessment. Do not treat 28% AIO coverage as an exact statement about all Google searches. |
| Phase 2 / Unavailable | No number exists. The slot is intentionally empty. | Do not estimate or assume. Flag as data gap in any model that references this dashboard. |
The following questions cannot be answered by this dashboard in its current Phase 1 state. If you need these numbers, they require a paid data source:
- How many people are actively using the Gemini app today? — Requires Sensor Tower or data.ai (app DAU/MAU). Not in Phase 1.
- What is Google's exact search revenue this quarter? — Requires the earnings filing. The dashboard provides directional proxies, not financial projections.
- What is Google's actual cost-per-click this quarter? — Keyword Planner bid ranges are planning-tool estimates. The actual blended CPC is in Google's 10-Q.
- How much traffic does google.com receive? — Requires a paid web-intelligence panel (SimilarWeb / Semrush). Not in Phase 1.
- Is AI Overview causing Google to lose revenue? — We can see AIO is expanding. We cannot yet directly measure the revenue impact without CTR and revenue-per-click data (Phase 2).
- What are Google's enterprise AI contract prices? — Only public list prices are shown. Enterprise rates are negotiated privately and not disclosed.
- google.com traffic estimates (MoM & YoY): Via approved paid web-intelligence subscription. Will add device split, geographic breakdown, and source-of-traffic.
- Gemini app engagement: Absolute DAU/MAU, download rankings, and session depth via Sensor Tower or data.ai. Will allow direct comparison to ChatGPT app.
- CTR / monetisation panels (Module 04): Click-through rates by vertical using live benchmark campaigns or paid Search Console access.
- Expanded AIO basket: 200 keywords (up from 120) with Finance 40%, Legal 20%, Travel 20%, Health 20% weighting to better represent high-monetisation queries.
Current basket: 120 keywords, manually curated, monthly spot-check. Composition:
- Informational queries: ~45% (e.g., "how does X work", "what is Y")
- Commercial queries: ~30% (e.g., "best X for Y", "X vs Y comparison")
- Transactional queries: ~15% (e.g., "buy X", "X price")
- Navigational queries: ~10% (branded terms)
Important bias note: The basket is currently skewed informational. AI Overview naturally appears more often on informational queries (general knowledge, explanations) than on commercial or transactional queries (buying decisions). This means our 28% overall coverage figure is higher than what you would see in a basket weighted toward high-value commercial queries. Phase 2 will rebalance the basket toward commercial/transactional to give a more conservative and monetisation-relevant penetration figure.
APPENDIX — PUBLISHED AI MODEL PRICING
Published AI Model Pricing Reference
Official public list prices · Google, OpenAI, Anthropic, Meta · does not include enterprise or negotiated rates
Published list
Vendor public pages
| Provider | Model | Input / 1M tokens | Output / 1M tokens | Context | Source |
|---|---|---|---|---|---|
| Google — Vertex AI / Gemini API | |||||
| Gemini 2.5 Pro Flagship | $1.25 | $10.00 | 1M tokens | Vertex Pricing | |
| Gemini 2.5 Flash | $0.15 | $0.60 | 1M tokens | Vertex Pricing | |
| Gemini 2.0 Flash | $0.10 | $0.40 | 1M tokens | Vertex Pricing | |
| Gemini 1.5 Flash | $0.075 | $0.30 | 1M tokens | Vertex Pricing | |
| OpenAI — API | |||||
| OpenAI | GPT-4.1 Latest | $2.00 | $8.00 | 1M tokens | OpenAI Pricing |
| OpenAI | GPT-4.1 mini | $0.40 | $1.60 | 1M tokens | OpenAI Pricing |
| OpenAI | GPT-4o | $2.50 | $10.00 | 128K tokens | OpenAI Pricing |
| OpenAI | o3 Reasoning | $10.00 | $40.00 | 200K tokens | OpenAI Pricing |
| OpenAI | o4-mini | $1.10 | $4.40 | 200K tokens | OpenAI Pricing |
| Anthropic — Claude API / AWS Bedrock | |||||
| Anthropic | Claude 3.7 Sonnet Latest | $3.00 | $15.00 | 200K tokens | Anthropic / Bedrock |
| Anthropic | Claude 3.5 Haiku ↓ Reduced | $0.80 | $4.00 | 200K tokens | Anthropic / Bedrock |
| Anthropic | Claude 3 Opus | $15.00 | $75.00 | 200K tokens | Anthropic / Bedrock |
| Meta — Llama (via hosted inference providers) | |||||
| Meta / hosted | Llama 4 Scout Latest | $0.17 | $0.17 | 10M tokens | Together / Groq list |
| Meta / hosted | Llama 4 Maverick | $0.27 | $0.85 | 1M tokens | Together / Groq list |
| Meta / hosted | Llama 3.3 70B | $0.59 | $0.79 | 128K tokens | Together / Groq list |
| Pending Public Pricing Announcement | |||||
| OpenAI | GPT-5 Not listed | TBA | TBA | TBA | No public page |
| Gemini 3.0 / Ultra Not listed | TBA | TBA | TBA | No public page | |