HubSpot AI lead attribution prompts for evidence-led reporting
Start with content strategy, expert editorial support, and a coordinated plan across owned pages, external content, paid campaigns, community, and social. Use the prompts below to translate that plan into documented review steps with named human owners.
HubSpot already labels many AI visits for you. Its AI Referrals source tags clicks from recognized platforms, including ChatGPT and Perplexity. The six prompts below build on that baseline. They help you attribute ChatGPT and Perplexity leads in HubSpot with a clean schema and a finance-ready story.
AirOps is an enterprise AEO (answer engine optimization) and content operations platform. AirOps experts set strategy and run execution across owned content, external content, AI ads, community, and social. Every prompt here keeps a person in charge of each HubSpot change.
- HubSpot files recognized AI platform visits as AI Referrals, with the platform domain in Drill-Down 1.
- Store extra AI evidence in a separate custom property instead of editing Original Traffic Source.
- Each prompt uses only data you paste and reports a minimum observed count.
- Every prompt ends with a HOLD gate, so a person reviews changes before HubSpot updates.
- No prompt guarantees AI citations or revenue, and referrer-less visits stay in Direct.
How do you attribute ChatGPT and Perplexity leads in HubSpot?
Start with HubSpot's native AI Referrals source. Then add your own evidence property and a review gate. HubSpot lists AI Referrals as a value in its traffic source properties guide.
Drill-Down 1 holds the AI platform domain, such as chatgpt.com. Drill-Down 2 holds the utm_campaign value when one exists. Associated companies and deals mirror these properties.
- Confirm AI Referrals shows in your Original and Latest Traffic Source reports.
- Add a custom evidence property for signals HubSpot does not classify.
- Capture landing-page query parameters in hidden form fields.
- Keep lifecycle stage and source in separate properties.
- Route every proposed change through a named reviewer.
These counts cover clicked visits that kept a referrer or an AI platform UTM (Urchin Tracking Module) tag. Report them as a minimum observed figure. Use the map below to pick a prompt.
| Prompt | You supply | AI returns | HOLD gate |
|---|---|---|---|
| 1. Property schema | Property export CSV and portal tier | Custom properties with locked internal names | Admin approves names |
| 2. Form and hidden-field plan | Form list and sample landing URLs | Hidden-field map per form | Web owner tests one form |
| 3. Lifecycle association rules | Lifecycle settings and workflow list | Forward-only rules and a deal copy plan | Ops lead approves each rule |
| 4. Dashboard narrative | Report exports and portal tier | Report list and header narrative | Marketing lead reviews wording |
| 5. Quality assurance (QA) for miscategorized traffic | Contact export with source fields | Flagged records with likely causes | Analyst reviews every flag |
| 6. Monthly chief financial officer (CFO) readout | HubSpot report export | One-page readout in minimum observed counts | CMO signs off |
Prompts 3 and 4 for lifecycle rules and the dashboard narrative
Lifecycle stage shows where a contact sits, and source shows where it came from. Keep them apart.
Use the lifecycle settings and stage order configured in your portal. Confirm which automations update stages before proposing new rules.
- Never reset lifecycle stage based on AI evidence.
- Copy ai_platform_observed to the deal at deal creation.
- Run stage changes and source changes in separate workflows.
Prompt 3 writes lifecycle association rules
Paste your stage settings and live workflows into this prompt.
You are a HubSpot lifecycle operations assistant. Write association rules for AI-referred contacts.
Inputs:
[PASTE: lifecycle stage settings and stage order]
[PASTE: active workflow names and triggers]
[PASTE: approved property schema from Prompt 1]
[YOUR: portal tier]
Rules:
- Use only the supplied settings. Never invent stages, workflows, or numbers.
- Lifecycle stage moves forward only. Never propose a backward move.
- Keep source properties and lifecycle stage in separate rules.
- Plan how AI evidence copies from contact to company and deal.
- Label any workflow step your tier does not support "verify in portal".
- Call any count a "minimum observed" figure.
Output:
1. Numbered rules, each with trigger, condition, and action.
2. Conflicts with existing workflows.
HOLD: a human ops lead approves each rule before anything changes in HubSpot.Prompt 4 drafts the dashboard narrative
HubSpot offers three attribution report types: Contact Create, Deal Create, and Deal Revenue Attribution. Its models are First Touch, Last Touch, Linear, Time Decay, and Empirical.
Deal-create and revenue attribution require Marketing Hub Enterprise. Empirical replaces the older U-shaped and W-shaped models. In these reports, "lead" means the contact.
Write the narrative for an executive reader. Our hub on AI search for marketing leaders covers that CMO view.
You are a B2B marketing analyst. Draft a dashboard narrative for AI Referrals in HubSpot.
Inputs:
[PASTE: report CSV exports or report descriptions]
[YOUR: portal tier]
[YOUR: attribution model in use: First Touch, Last Touch, Linear, Time Decay, or Empirical]
Rules:
- Use only the supplied data. Never invent numbers, trends, or causes.
- Call every AI Referrals figure a "minimum observed" count.
- Name the attribution model behind each attribution figure.
- Skip deal-create and revenue attribution unless the tier is Marketing Hub Enterprise.
- Note that "lead" means the contact in attribution reports.
- Never promise future citations or revenue.
Output:
1. A report list with name, object, filters, and Drill-Down 1 breakdown.
2. A 150-word narrative for the dashboard header.
HOLD: a human marketing lead reviews the narrative before anything changes in HubSpot.Prompts 5 and 6 to QA AI referral traffic and brief your CFO
HubSpot checks its traffic source rules in a fixed order. Social, paid search, paid social, and email rules run before AI Referrals. So a chatgpt.com click tagged utm_medium=cpc lands in paid search.
Recognized AI platforms include ChatGPT, Claude, Perplexity, Microsoft Copilot, Gemini, Meta AI, Mistral, Poe, and Grok. A visit with no referrer and no tracking URL becomes Direct traffic.
- Check UTM tags on links you control for cpc or email values.
- Confirm the HubSpot tracking code runs on every form page.
- Review cookie consent settings and ad-blocker effects on tracking.
Prompt 5 runs QA for miscategorized traffic
QA here means finding AI visits HubSpot filed elsewhere.
| Symptom | Likely cause | Check |
|---|---|---|
| chatgpt.com visit filed as paid search | cpc tag or gclid on the link | Review tags on links you control |
| AI visit filed as email | Email rule runs before AI Referrals | Look for email UTM values on the landing URL |
| Form fill with a Direct source | Form page lacks the tracking code | Confirm the tracking code on that page |
| Contact with no source detail | Ad-blocker or cookie consent blocked tracking | Compare against your consent settings |
| ChatGPT click with no tag or referrer | Referrer and utm_source both missing | Leave it in Direct without new evidence |
Attach a contact export to this prompt.
You are a HubSpot data QA analyst. Flag AI visits that HubSpot filed under another source.
Inputs:
[PASTE: contact export with Original Traffic Source, Drill-Down 1, Drill-Down 2, First Page Seen, and create date]
[PASTE: UTM values used in your own campaigns]
[YOUR: date range]
Rules:
- Use only the supplied rows. Never invent records, numbers, or causes.
- Apply HubSpot's rule order. Social, paid search, paid social, and email run before AI Referrals.
- Flag rows outside AI Referrals that show chatgpt.com or perplexity.ai in any field.
- Never reassign Direct rows without a matching UTM or referrer.
- Report flagged totals as "minimum observed".
- Recommend fixes only for links you control.
Output:
1. A table with record ID, current source, evidence found, and likely cause.
2. A fix list for your own campaign tags.
HOLD: a human analyst reviews every flag before anything changes in HubSpot.Prompt 6 builds the monthly CFO readout
Your CFO needs figures that hold up under questions in a budget review. Map each readout field to a HubSpot source first.
| Readout field | HubSpot source | How to phrase it |
|---|---|---|
| New AI-referred contacts | Contact report filtered on Original Traffic Source | Minimum observed new contacts |
| Platform split | Drill-Down 1 | Counts by domain, such as chatgpt.com |
| Extra evidence | ai_referral_evidence property | Signals HubSpot did not classify |
| Deals created | Deal Create Attribution on Marketing Hub Enterprise | Model named next to each figure |
| Unmeasured visits | Direct and unmatched rows | Listed as unmeasured with no estimate |
HubSpot supports CSV and Excel report exports. Single-object exports include Selected properties; advanced custom reports use fields from Configure. Confirm the exported columns against the HubSpot export guide.
You are a finance-minded marketing analyst. Write a monthly CFO readout from a HubSpot export.
Inputs:
[PASTE: HubSpot report export, CSV or Excel]
[YOUR: portal tier]
[YOUR: attribution model used in the export]
Rules:
- Use only values present in the export. Never invent, round up, or extrapolate numbers.
- Label every AI figure "minimum observed".
- Name the attribution model next to each pipeline figure.
- Omit revenue attribution unless the export comes from Marketing Hub Enterprise.
- List fields missing from the export as open questions.
- Never claim AI search caused revenue. Never promise future citations.
Output:
1. Five bullets for the CFO, each under 20 words.
2. A table of figures with source report names.
3. Open questions and data gaps.
HOLD: a human CMO reviews the readout before anything changes in HubSpot or leaves it.No prompt here guarantees AI citations or revenue. Each one gives you a defensible minimum that a person has reviewed. Pair the readout with outcome-based AEO measurement for the wider picture.
Key takeaways
- AI Referrals is your HubSpot baseline, so build on it instead of rebuilding it.
- Lock internal names before you create a single property.
- Keep AI evidence out of Original Traffic Source and apart from lifecycle stage.
- Rule order explains many misfiled AI visits, and Direct visits stay unrecoverable.
- Minimum observed beats maximum imagined, so name the model and gate every change.
FAQ about HubSpot AI referral attribution
Does HubSpot track ChatGPT and Perplexity leads on its own?
Yes, for clicks HubSpot can recognize. ChatGPT (chatgpt.com) and Perplexity (perplexity.ai) are recognized AI platforms filed as AI Referrals. Visits without a referrer or AI tag still land in Direct.
Why do AI leads show up as Direct traffic in HubSpot?
HubSpot files a visit as Direct when the referring information is empty or lost. Its direct traffic explainer also names forms on pages without the tracking code. No prompt recovers those visits.
Should I overwrite Original Traffic Source with an AI value?
Preserve Original Traffic Source and store added AI evidence in a separate custom property. HubSpot documents that editing the source clears its drill-down and First Referring Site values.
Which HubSpot tier do I need for AI revenue attribution?
Confirm each requirement in your own portal, since HubSpot updates these pages often.
- Marketing Hub Enterprise for deal-create and revenue attribution.
- Marketing Hub or Content Hub Professional or Enterprise for the traffic analytics tool.
- Professional or Enterprise for workflow edits to traffic source values.
Do these prompts guarantee AI citations or pipeline?
Use these prompts to organize existing evidence, not to forecast citations or pipeline. Review source quality, reporting scope, and uncertainty before presenting results.
AirOps for HubSpot AI lead attribution
The six prompts give you a measurement system your CFO can question and trust.
- A schema and hidden-field plan that protect HubSpot's native source data.
- Lifecycle rules and QA checks that keep AI evidence clean.
- A CFO readout built on minimum observed counts behind HOLD gates.
AirOps works alongside you as a partner on that program. The AirOps platform gives you intelligence on where your brand stands in AI search. AirOps experts then run owned content, external content, AI ads, community, and social against the pipeline you track.