1. Overview

The Intellectica AI Discourse Index™ measures the composition of public AI conversation in a country, expressed as a percentage breakdown across five mutually exclusive dimensions: Curiosity & Learning, Tool Adoption, Career & Opportunity, Anxiety & Risk, and Ethics & Governance.

The index is published weekly. Each release produces a snapshot such as:

Composition shifts over time as cultural, technological, and regulatory events reshape what people search for. Tracking these shifts reveals the character of a country's AI engagement — not whether it uses AI, but how it thinks about AI.

The pilot edition focuses on Greece. The methodology is designed to be replicable for any country. The companion Intellectica AI Pulse Index™ measures absolute AI absorption levels; the Discourse Index complements it by revealing the texture of the conversation beneath those levels.

1.1 Why weekly cadence

Discourse composition shifts faster than structural absorption. When a significant event occurs — a major model launch, regulatory news, an AI-related controversy — the composition of what people search shifts within days, not months. A monthly cadence would average away these shifts; a weekly cadence captures them.

Google Trends data is natively weekly when using custom date ranges of more than 90 days. The weekly cadence therefore matches the natural granularity of the underlying data source, avoiding interpolation artifacts.

1.2 Why this index exists

Existing AI indices focus on adoption levels, capability rankings, or investment volume. None measure the texture of public conversation: what kind of engagement does a country have with AI? Is it dominated by curiosity, practical tool use, career aspiration, fear, or institutional debate?

The texture of conversation often precedes behaviour change. Rising curiosity tends to lead adoption. Rising anxiety tends to precede regulatory action. The Discourse Index provides a leading indicator of where a country's AI relationship is heading.

2. Data Source

The index relies on a single data source: Google Trends. This is intentional. Google searches are uniquely well-suited to measuring public discourse composition because they combine four properties:

• Emotional honesty. People search what they actually wonder about, fear, want, or need.
• Intent signal. Unlike news mentions or social posts, a search is a deliberate act by a person seeking something specific.
• Broad coverage. Google handles the vast majority of search traffic in Greece across all demographics.
• Public availability. No paid APIs, no proprietary datasets — the methodology can be independently replicated.

3. The Five Dimensions

The discourse is decomposed into five mutually exclusive dimensions. Each represents a distinct mode of public engagement with AI:

3.1 Keyword basket selection principles

Each dimension uses a basket of keywords combined with the Google Trends "+" operator into a single search query. Keywords were selected using four principles:

• Longevity test. Brand-specific terms were excluded; generic concept names were preferred.
• Low degrees of freedom. Short standardised phrases that many people would type identically.
• Authentic intent. Each keyword had to plausibly express the dimension's sentiment.
• Empirical validation. Every candidate was tested in Google Trends; keywords with flat-zero signal were eliminated.

3.2 Dimension 1: Curiosity & Learning

Measures the share of AI discourse driven by people seeking to understand AI concepts — the entry point of engagement.

3.3 Dimension 2: Tool Adoption & Productive Use

Measures the share of AI discourse driven by people actively using AI for productive purposes.

3.4 Dimension 3: Career & Opportunity

Measures the share of AI discourse driven by people viewing AI as a professional or entrepreneurial opportunity.

3.5 Dimension 4: Anxiety & Risk

Measures the share of AI discourse driven by fear, concern, and risk awareness.

3.6 Dimension 5: Ethics & Governance

Measures the share of AI discourse driven by institutional, regulatory, and ethical debate.

4. Calculation Formula

Each weekly publication is calculated through a single Google Trends Compare query followed by simple arithmetic. The single-query design ensures all five dimensions share a common scale, enabling valid percentage composition.

Step 1: Run the master query (weekly)

A single Google Trends Compare query with all five dimension baskets entered as separate slots. Custom date range from January 2026 through the current week (Greece, Web Search). Ranges longer than 90 days produce weekly granularity automatically.

Step 2: Extract latest-week values

For each of the five trend lines, record the value of the latest complete week: V₁ through V₅. These are raw Google Trends values on the shared 0–100 scale.

Step 3: Apply anchor scaling

Step 4: Calculate weekly composition percentages

Step 5: Calculate 4-week rolling average

The Rolling_Share is the official index value shown in charts and headline reports. The Weekly_Share is preserved as a data point for archival and intra-week analysis.

5. Time-Series Continuity

Google Trends scores are normalised within each query window. If the time range expands week-by-week, raw values shift even when underlying interest is stable. The anchor protocol solves this by fixing a permanent reference point.

5.1 Reference period selection

The reference period was selected empirically by examining the full historical record of Greek AI search data. Selection criteria:

• All dimensions must have stable, non-near-zero signal. Otherwise the scale factor calculation becomes mathematically unstable.
• The period should not be an extreme high or low. Anchoring to a peak creates one-directional drift.
• The period should represent a stable plateau, not a transition. Major events that shift composition should be excluded from the anchor period.

Selected reference period

Reference period: Average of February 2026 and March 2026. This two-month average represents the post-stabilisation baseline. All five dimensions had reached stable levels after the October 2025 structural break.

These anchor values are immutable for the duration of methodology version 1.x.

5.2 Anchor protocol (weekly application)

1. Initial measurement (one-time): Calculate the Feb–Mar 2026 averages from weekly data; these become V_{1_anchor} through V_{5_anchor}.
2. Weekly query: Each week, run the master query with a custom date range from January 2026 through the current week.
3. Scale factor calculation: Compute the Feb–Mar 2026 average in the current query (V_{i_ref}) and calculate Scale_i = V_{i_anchor} / V_{i_ref}.
4. Apply scale: Multiply the current week's raw value by the dimension's scale factor before computing percentages.

5.3 Re-anchoring

Re-anchoring is triggered only by sustained shifts, not single-week spikes:

• All-time high exceeding the original anchor by more than 50%, sustained for 4+ consecutive weeks
• Structural break in discourse composition similar to the October 2025 event
• Geographic expansion requiring a new reference geography

When re-anchoring occurs, both old and new anchor values are published, and the prior 8 weeks are recalculated under the new anchor. Re-anchoring is a major version event.

6. Auxiliary Metrics

Beyond the five dimension shares, three derived metrics are reported each week.

6.1 Adoption-to-Anxiety Ratio

The headline metric — a single number capturing emotional balance.

Reported as both a weekly value and a 4-week rolling average.

6.2 Discourse Diversity Index

Higher values indicate balanced discourse; lower values indicate concentration in a single dimension. A five-way even split produces 0.80; complete concentration produces 0.

6.3 Week-over-Week Composition Shifts

Using the smoothed rolling share rather than raw weekly values prevents single-week noise from triggering false alerts.

7. Publication Format

7.1 Weekly publication

Each weekly release includes:

• Composition snapshot: the five dimension percentages (4-week rolling), displayed as a stacked bar
• Headline metric: the Adoption-to-Anxiety Ratio (rolling 4-week)
• Week-over-week movers: the dimensions with the largest shifts
• Brief contextual note: 1–2 sentence summary connecting movements to identifiable events

7.2 Monthly deep-dive companion

Once per month, a longer report aggregates weekly data with additional analysis: composition trend charts, Adoption-to-Anxiety trajectory, event correlation, and analyst commentary.

8. Limitations & Considerations

• Search-based discourse only. Behaviours that bypass search (social media, direct app usage, private channels) are invisible.
• Anxiety dimension is structurally weak in Greece. A low Anxiety percentage may indicate concern is expressed through channels other than search, not that concern is absent.
• Pre-October 2025 backtest is not reliable. Three of the five dimensions had insufficient signal before late October 2025. The index begins meaningful operation from this structural break point forward.
• Weekly data is noisier than monthly. Mitigated through the 4-week rolling average used as the official reported figure.
• Keywords were calibrated for Greece in 2026. Cultural and linguistic patterns shift. Periodic recalibration is required (see Section 9).
• The index measures relative composition, not absolute volume. A rising Curiosity share could mean curiosity grew or other dimensions shrank. The companion Pulse Index complements this.
• Five dimensions are not exhaustive. Rich sub-textures within each dimension are not captured at this level of decomposition.

9. Versioning & Methodology Updates

9.1 Versioning convention

• v1.x (minor versions): keyword additions or substitutions within a basket, documentation refinements. No structural impact on comparability.
• v2.0+ (major versions): addition or removal of dimensions, structural formula changes, basket redefinition, or re-anchoring.

9.2 Keyword basket updates

Keyword baskets are reviewed every six months. Triggers for change:

• A keyword shows decaying signal over 8+ consecutive weeks
• A new search pattern emerges with continuous high signal fitting an existing dimension
• Cultural or technological shifts make a keyword obsolete

Minor updates (single keyword swap) are applied silently and logged. Major overhauls (3+ keywords) trigger a major version bump with backward recalculation.

9.3 Re-anchoring events

Re-anchoring is a major version event, triggered only by sustained patterns (4+ consecutive weeks). Both old and new anchor values are published, and the prior 8 weeks are recalculated.

9.4 Major version communication

Major changes are communicated: "Starting this week, the AI Discourse Index methodology has been updated. Specifically: [list of changes]. The previous 8 weeks have been recalculated under the new methodology for continuity. Full methodology change notes are available at [link]."

10. Roadmap

• Cross-sectional version: comparing discourse composition across countries (requires language-neutral keyword baskets)
• Sub-dimension breakdowns within each major dimension
• Integration with the Intellectica AI Pulse Index for a combined dashboard
• Expansion to additional countries: Cyprus, Italy, Portugal, Spain
• Automated weekly data pipeline (pending Google Trends API access)
• Annual review of keyword baskets based on emerging search patterns