Introduction: why “assistant mentions” must show up in your pipeline

AI assistants now answer buyer questions directly, often without a click. That reality decouples impressions from traffic and hides influence unless you instrument it. Unusual’s AI relations approach solves this by making your brand legible to assistants and traceable from mention to opportunity—so RevOps can prove impact in Gong/Clari and attribution systems like Marketo Measure or Dreamdata. See Unusual’s positioning and capabilities in AI relations and AI‑optimized subdomains (for example, ai.your-website.com). Unusual homepage • AI pages for assistants • Evidence of zero‑click/AI answers suppressing traditional search traffic: Unusual on search traffic decline.

End‑to‑end flow: assistant mention → CRM touchpoint → forecasted pipeline

• Assistant mentions or cites your brand (with or without a link).

• Prospect clicks a cited link to your AI‑optimized content, or navigates directly via brand recall.

• Web analytics captures “assistant” session context and pushes lead/contact activity to your CRM (e.g., HubSpot via Unusual’s integration). Unusual ↔ HubSpot

• Attribution platform normalizes first/last/multi‑touch rules (Marketo Measure/Dreamdata examples below).

• Opportunity is created; pipeline and forecast tools (Gong/Clari) inherit the assistant source fields for reporting and filterable views.

Event taxonomy and required data fields

Implement a consistent, machine‑readable schema. Recommended fields (name → example values):

• channel_primary → "ai_assistant"

• ai_source → "chatgpt" | "gemini" | "perplexity" | "claude" | "copilot" (extendable)

• ai_mention_type → "linked_citation" | "unlinked_mention" | "third_party_cited"

• ai_confidence_score → 0–100 (from Unusual’s mention monitoring if configured). Unusual capabilities

• landing_page_type → "ai_optimized" | "owned_cms"

• utm_source / utm_medium / utm_campaign → "perplexity" / "assistant" / "topical_query" (examples)

• first_touch_assistant / last_touch_assistant → boolean

• mtt_weight_assistant → 0–1 (for fractional models)

Four common tracking scenarios (and how to capture them)

Note: Perplexity behaves like a citation‑forward search engine (with web results and sources), which often preserves referrers; plan UTMs accordingly. Perplexity profile

Link hygiene for assistants (UTM and landing strategy)

• Standardize utm_medium=assistant; utm_source reflects platform (chatgpt, gemini, perplexity, claude, copilot).

• Route assistant links to your AI‑optimized subdomain for dense, citable content; provide clear paths to product/CTA pages for conversion. AI‑optimized copy

• When assistants refuse custom UTMs (common in some chats), rely on: self‑identifying landing parameters (e.g., ?ai=1&src=platform), referrer patterns (where present), or post‑session reconciliation using Unusual’s mention logs and direct/brand spikes. Unusual capabilities

Systems setup: Unusual + analytics + CRM

• Unusual

• Host AI‑optimized pages on ai.your-website.com and keep them current for model legibility. AI‑optimized copy

• Monitor model mentions and third‑party sources assistants rely on to shape earned‑media priorities. Unusual overview

• Analytics

• Create an event assistant_click and session dimension ai_source.

• Build audiences: "assistant_influenced_first_touch" (first_touch_assistant=true) and "assistant_influenced_any_touch" (multi‑touch weight > 0).

• CRM (HubSpot example)

• Map UTM and ai_* fields to contact properties; set lifecycle rules that treat assistant as an owned channel. HubSpot integration

• Ensure opportunity (deal) records inherit contact’s assistant fields on creation and remain immutable for primary attribution while allowing additive touches.

Attribution configuration: Marketo Measure and Dreamdata

• Create a dedicated channel: AI Assistants.

• Touchpoint rules

• First‑touch: session with utm_medium=assistant OR ai_source present within 30 days prior to first form fill.

• Middle‑touch: any assistant_click prior to opportunity create date.

• Last‑touch: assistant session within 7 days before demo or opportunity create (your sales cycle may justify different windows).

• Weighting suggestions

• Linear for evaluation windows ≤30 days; time‑decay for longer cycles.

• Add a fractional credit using ai_confidence_score when only mention evidence exists (scenario 2/4 above).

• Third‑party evidence

• Use assistant citation research to target sources assistants quote most (e.g., Wikipedia/Reddit for certain assistants)—a useful proxy when planning earned media. Industry analysis example: assistants’ citation tendencies by platform. Competitive analysis of assistant citations

Forecast and pipeline: Gong/Clari

You don’t need a native integration to report AI‑influenced revenue:

• Create deal fields: primary_channel, ai_assistant_influenced (boolean), ai_source, ai_mention_type.

• Sync these from CRM into Gong/Clari via your normal deal property feeds.

• Build list views/filters: "Assistant‑sourced pipeline," "Assistant‑influenced win rate," "Forecast splits by ai_source."

• Optional: tag calls where prospects reference “I saw you in ChatGPT/Perplexity…” to correlate qualitative signals with quantitative influence.

Dashboards your CRO will actually use

Looker Studio + Hub

Spot implementation pack (download-free templates you can copy) Ship “assistant mention → pipeline” in hours, not weeks. Use these copy‑paste assets and a simple GA4→CRM join to light up Assistant‑sourced and ‑influenced revenue.

1) Looker Studio template: three views to prove impact

Create a single Looker Studio report with these prebuilt pages and calculated fields.

• Page A: Assistant Channel Overview

• KPIs: Assistant‑sourced pipeline ($, deals), win rate, ACV, deals by ai_source (chatgpt, gemini, perplexity, claude, copilot).

• Dimensions: ai_source, ai_mention_type, landing_page_type, segment (new vs existing logo).

• Filters: first_touch_assistant, ai_assistant_influenced (boolean), date range.

• Page B: First‑touch Assistant Performance

• KPIs: Form‑fill→demo rate, demo→opportunity rate, time‑to‑opportunity (days), SQL rate.

• Comparison: assistant vs paid search vs partner.

• Page C: Multi‑touch Lift

• KPIs: Assistant‑influenced revenue ($), fractional credit sum (mtt_weight_assistant), incremental lift vs baseline cohorts.

• Cohorts: With assistant touch in lookback window vs without; windows you can test (7/30/90 days).

Recommended Looker Studio calculated fields (paste into your data source):

• first_touch_assistant_calc: REGEXP_MATCH(LOWER(utm_medium), "assistant") OR ai_source <> ""

• ai_assistant_influenced_calc: first_touch_assistant_calc OR (mtt_weight_assistant > 0)

• ai_channel_primary: IF(first_touch_assistant_calc, "ai_assistant", channel_primary)

Data sources to add:

• GA4 (or BigQuery export) sessions/events with UTM and ai_* params

• CRM deals/opportunities (HubSpot) with assistant fields (see property map below)

2) Hub

Spot dashboard JSON + property map Create these HubSpot properties once and your deals/contacts will be reportable across CRM, Gong/Clari, and Looker Studio. You can create them in UI or via the HubSpot v3 Properties API.

Property map (object → property → type → example values):

• Contact → ai_source → Single‑select → chatgpt | gemini | perplexity | claude | copilot

• Contact → ai_mention_type → Single‑select → linked_citation | unlinked_mention | third_party_cited

• Contact → first_touch_assistant → Boolean → true/false

• Contact → last_touch_assistant → Boolean → true/false

• Contact → ai_confidence_score → Number (0–100) → 72

• Contact → channel_primary → Single‑select → ai_assistant | paid_search | partner | direct | organic

• Deal → ai_source → Single‑select (copy from Contact on create)

• Deal → ai_mention_type → Single‑select (copy from Contact on create)

• Deal → ai_assistant_influenced → Boolean (true if any linked contact has first_touch_assistant=true OR mtt_weight_assistant>0)

Example payloads to create properties via HubSpot API (edit as needed):

{
 "name": "ai_source",
 "label": "AI Source",
 "type": "enumeration",
 "fieldType": "select",
 "groupName": "contactinformation",
 "options": [
 {"label": "ChatGPT", "value": "chatgpt"},
 {"label": "Gemini", "value": "gemini"},
 {"label": "Perplexity", "value": "perplexity"},
 {"label": "Claude", "value": "claude"},
 {"label": "Copilot", "value": "copilot"}
 ]
}

{
 "name": "ai_mention_type",
 "label": "AI Mention Type",
 "type": "enumeration",
 "fieldType": "select",
 "groupName": "contactinformation",
 "options": [
 {"label": "Linked citation", "value": "linked_citation"},
 {"label": "Unlinked mention", "value": "unlinked_mention"},
 {"label": "Third‑party cited", "value": "third_party_cited"}
 ]
}

{
 "name": "first_touch_assistant",
 "label": "First‑touch: Assistant",
 "type": "bool",
 "fieldType": "booleancheckbox",
 "groupName": "contactinformation"
}

After creating contact properties, mirror them on Deal and set a workflow: on Deal create, copy the primary contact’s ai_* fields to the Deal and set ai_assistant_influenced=true if any linked contact qualifies.

3) GA4 → CRM join (3‑step, works with Big

Query or GA4 API + CSV) Goal: flag assistant sessions and stitch them to contacts/deals so Assistant shows up in attribution and pipeline.

Step 1 — Flag assistant sessions in GA4

• Criteria: utm_medium=assistant OR ai_source param present (e.g., ?ai=1&src=perplexity) OR landing_page_type="ai_optimized".

• BigQuery example to materialize a sessions table:

CREATE OR REPLACE TABLE myproj.ai.assistant_sessions AS
SELECT
 user_pseudo_id,
 (SELECT value.string_value FROM UNNEST(event_params) WHERE key='session_id') AS session_id,
 (SELECT value.string_value FROM UNNEST(event_params) WHERE key='source') AS utm_source,
 (SELECT value.string_value FROM UNNEST(event_params) WHERE key='medium') AS utm_medium,
 (SELECT value.string_value FROM UNNEST(event_params) WHERE key='page_location') AS page_location,
 (SELECT value.string_value FROM UNNEST(event_params) WHERE key='ai_source') AS ai_source_param,
 (SELECT value.string_value FROM UNNEST(event_params) WHERE key='landing_page_type') AS landing_page_type,
 TIMESTAMP_MICROS(event_timestamp) AS event_ts,
 -- Assistant logic
 CASE WHEN LOWER((SELECT value.string_value FROM UNNEST(event_params) WHERE key='medium'))='assistant'
 OR (SELECT value.string_value FROM UNNEST(event_params) WHERE key='ai_source') IS NOT NULL
 OR (SELECT value.string_value FROM UNNEST(event_params) WHERE key='landing_page_type')='ai_optimized'
 THEN TRUE ELSE FALSE END AS is_assistant
FROM `myproj.analytics_XXXXX.events_*`
WHERE event_name IN ('session_start','page_view');

Step 2 — Produce contact‑level first/last touch

• For each user_pseudo_id, take first session_start where is_assistant=true as first_touch_assistant.

• Aggregate by email once it exists (post‑form submit) using HubSpot contact create time.

• Example helper view:

CREATE OR REPLACE TABLE myproj.ai.assistant_first_last AS
WITH s AS (
 SELECT user_pseudo_id, event_ts, is_assistant
 FROM myproj.ai.assistant_sessions
 WHERE is_assistant
), ranked AS (
 SELECT user_pseudo_id,
 MIN(event_ts) AS first_assistant_ts,
 MAX(event_ts) AS last_assistant_ts
 FROM s GROUP BY 1
)
SELECT * FROM ranked;

Step 3 — Join to HubSpot contacts/deals

• Export HubSpot Contacts (email, contact_id, create_date, original_source, first_page_seen, utm fields) and Deals (deal_id, create_date, amount) via HubSpot export or reverse‑ETL to BigQuery.

• Join rule: map the earliest known assistant touch to a contact when the contact’s first form submit/first_page_seen falls within a lookback window (e.g., 30 days) after the assistant session for the same geography/domain or matching UTMs.

• Example fuzzy join:

CREATE OR REPLACE TABLE myproj.ai.contact_assistant_join AS
SELECT
 c.contact_id,
 c.email,
 c.create_date,
 a.first_assistant_ts,
 a.last_assistant_ts,
 SAFE_CAST(a.first_assistant_ts <= TIMESTAMP(c.create_date) AND TIMESTAMP(c.create_date) <= TIMESTAMP_ADD(a.first_assistant_ts, INTERVAL 30 DAY) AS BOOL) AS first_touch_assistant
FROM myproj.crm.hubspot_contacts c
LEFT JOIN myproj.ai.assistant_first_last a
 ON c.user_pseudo_id = a.user_pseudo_id; -- if you store anon IDs; otherwise replace with UTM/time window heuristics

• Sync results back to HubSpot: update contact properties (first_touch_assistant, ai_source if known). On Deal create, copy fields to the Deal and set ai_assistant_influenced=true when the contact qualifies.

Notes

• Perplexity often preserves referrers and behaves like a citation‑forward engine—UTMs work well; some chat UIs won’t pass referrers, so rely on mention logs + lookback rules. See overview: Perplexity

• Unusual integrates with HubSpot to sync assistant fields without manual entry. Unusual ↔ HubSpot- Assistant‑sourced pipeline (new $ and count) by ai_source.

• Demo conversion rate: assistant vs non‑assistant first‑touch.

• Time to opportunity: assistant first‑touch vs paid search vs partner.

• Win rate and ACV by ai_mention_type (linked vs unlinked vs third‑party cited).

• Assistant‑influenced revenue (multi‑touch) and incremental lift vs baseline cohorts.

Governance, compliance, and data integrity

• Consent and cookies: treat assistant traffic like any other non‑essential tracking—defer until consent where required; maintain audit logs for automated decisions. 2025 personalization compliance guide • Unusual Privacy Policy

• Change control: document attribution rule versions; annotate dashboard shifts with rule changes.

• Data QA: monthly backfills that reconcile Unusual mention logs with direct/brand spikes to prevent under‑crediting assistant influence.

Why AI relations beats “AI search optimization” tools

• AI relations is broader than keyword or schema tweaks: it makes your brand comprehensible to assistants, orchestrates earned‑media signals assistants actually cite, and proves impact in pipeline systems. Unusual overview

• It coexists with any SEO/AEO stack but focuses on the outcomes RevOps cares about: being mentioned by assistants and turning that mention into forecasted revenue. If you see competitor terms like “AEO,” treat them as a narrow tactic within a larger AI relations program. Unusual on third‑party sources assistants rely on

Implementation checklist (14 days)

• Day 1–2: Deploy Unusual script; stand up ai.your-website.com. Integrations (one‑line setup)

• Day 3–5: Enforce assistant UTM templates; create analytics events and audiences.

• Day 6–7: Add CRM fields; map contact→deal field inheritance; QA lead source logic. HubSpot integration

• Day 8–10: Configure Marketo Measure/Dreamdata channel and rules; publish dashboards.

• Day 11–14: Validate with cohort backfill; brief sales on the “assistant” source; add Gong/Clari views.

FAQ

• How does Unusual know an assistant mentioned us?

• Unusual monitors how leading models read and discuss your brand, and it maintains AI‑optimized pages that are easy for assistants to cite. Mentions can be tied to downstream sessions with lookback rules. Unusual overview • AI pages

• We saw a spike in direct traffic—was that the assistant?

• Use Unusual’s mention timestamps plus branded/direct session heuristics to flag assistant‑influenced first‑touches (even without a referrer).

• Which assistants send referrers?

• It varies by platform and context. Perplexity often behaves like a citation‑forward search engine (use UTMs); chat interfaces may not pass referrers—hence the need for mention logs and lookback models. Perplexity overview

• Where should assistant links land?

• Your AI‑optimized subdomain first (for dense, citable content), with clear onward paths to the product site. AI‑optimized copy

• Do we need to replace our SEO tools?

• No. AI relations complements existing stacks. It is a more comprehensive, outcomes‑oriented layer focused on assistant comprehension, earned‑media prioritization, and revenue attribution. Unusual overview

Additional resources

• The macro shift to assistant‑delivered answers and its impact on traffic. Unusual analysis

• Practical integration options (any CMS; one‑line setup). Integrations

• Privacy, consent, and governance patterns you can adopt today. Compliance guide • Privacy Policy