Book A Demo
Skip to the ideas

A growing field guide

Marketing ideas
worth testing.

Practical ways to use AI across sales, creative, websites and search. Find an idea, shape a small test, and measure what actually changes.

Explore the ideas

The repository

Choose your next experiment.

S and A are editorial priorities, not measured performance. S marks the boldest starting points; A marks supporting tests. The 30-reel, 100-page and 1,000-question counts are proposed scales, not results or requirements. Start small and validate with real buyers.

Showing 12 of 12 ideas

S tierSales

Prospect funnel rebuild

Rebuild a prospect's funnel before the first call. Open with a working mockup of a better buying journey instead of another pitch.

Plan the first test
First test
Choose one prospect. Review its public offer, page and next step, then build a private, clearly labeled concept using original or authorized assets. Walk through the journey on the call.
What to measure
Track qualified next meetings and progression to an agreed proposal. Compare with similar pitch-led conversations, not every deal in the pipeline.
Keep in view
A mockup is not the prospect's live website. Keep it private, label assumptions and leave its forms, customer data and systems untouched.
Link to this idea
S tierVideo & creative

Trial-reel tournament

Turn one raw recording into 30 reels with different hooks, structures and lengths. Test with non-followers through trial reels, then share the winners with followers.

Plan the first test
First test
Use footage you own or have permission to edit. Map the hook, pacing and structure of strong reference videos, then create original variations. Confirm trial-reel access on the account before planning the test.
What to measure
Compare retention, qualified profile visits and relevant inquiries over a consistent observation window. Promote the strongest versions, not just the one with the most views.
Keep in view
Thirty is a proposed test scale. Start smaller if production capacity is limited. Borrow format principles, not another creator's footage, music or distinctive assets.
Link to this idea
S tierPaid media

Competitor ad library scan

Review competitor ads daily where public libraries allow it. Record the hook, offer, format and first-seen date to spot patterns worth testing.

Plan the first test
First test
Pick a small competitor set and the relevant public ad libraries. Save links and a dated observation for each visible ad. Separate a new creative from a new offer.
What to measure
Track which observations lead to a distinct test and what that test does to qualified lead cost or conversion quality in your own account.
Keep in view
Public libraries do not expose every ad or its performance. First seen is not necessarily launch date, and a long-running ad is not proof that it works.
Link to this idea
S tierAI search

Canary offers

Place a distinct, trackable offer on pages AI tools already cite. Look for evidence that the offer reaches buyers through those answers.

Plan the first test
First test
Confirm a citation with a saved question, answer and date. Add a genuine offer with a dedicated destination on a page you control, then repeat the question and review visits and inquiries.
What to measure
Record whether the offer appears in sampled answers, whether visitors use its destination, and whether leads mention it. Keep confirmed and unknown sources separate.
Keep in view
AI referrals can lose tracking information. An offer appearing in an answer does not establish that it caused a purchase.
Link to this idea
S tierAI search

Daily buyer-question panel

Build a repeatable panel of 1,000 buyer questions and sample AI answers daily. Flag changes in brand mentions, recommendations and competitor citations.

Plan the first test
First test
Start with a smaller set of real buyer questions grouped by buying stage. Use permitted interfaces, record the tool and answer date, and review changed answers before expanding toward 1,000 questions.
What to measure
Compare mention and citation share within the same documented question set. Investigate repeated changes before acting on a single answer.
Keep in view
This is a proposed research scale, not a claim of exhaustive monitoring. Answers vary by session, location and tool; access, cost and rate limits constrain coverage.
Link to this idea
S tierAI search

Citation-source takeover

Map the pages AI tools cite in your niche. Identify relevant owners and pitch a useful reason for your brand to be included.

Plan the first test
First test
Sample high-intent questions, collect the cited URLs and review each page. Shortlist owners whose readers would benefit from an original example, resource or expert contribution.
What to measure
Track relevant placements, referral visits, qualified inquiries and changes in sampled citations. Keep outreach acceptance separate from AI visibility.
Keep in view
No sample finds every cited page. Owners retain editorial control. Avoid spam, misleading endorsements and paid links disguised as independent coverage.
Link to this idea
A tierVideo & creative

Format reverse-engineering

Turn standout reels in your niche into a weekly template library: hook type, cut pace, caption style and length.

Plan the first test
First test
Choose a few relevant reels and annotate how each earns attention. Translate the pattern into a reusable outline, then make a new video with your own message and authorized footage.
What to measure
Compare production time, retention and qualified actions across your original videos using the templates. Keep the audience and topic differences visible.
Keep in view
Visible popularity is a clue, not evidence of sales. Study structure without copying assets, impersonating a creator or reproducing protected material.
Link to this idea
A tierWebsite & conversion

100 funnel variants, pre-screened

Build 100 page variations, use AI buyer personas to narrow the options, then give real traffic to the survivors.

Plan the first test
First test
Vary one clear hypothesis at a time, such as the offer, headline or proof order. Use AI personas to surface objections and obvious gaps, then have a marketer select a small set for a controlled live test.
What to measure
Judge the live finalists on qualified conversion rate and downstream lead quality. Document traffic allocation and the observation window before starting.
Keep in view
AI-persona scores are simulated feedback, not buyer evidence. One hundred is an exploration scale, not a reason to split limited traffic across 100 live pages.
Link to this idea
A tierWebsite & conversion

Segment-personalized pages

Adapt the headline, proof and next step to the visitor's industry, referrer or known deal stage.

Plan the first test
First test
Start with two explicit segments and approved proof for each. Use a reliable, privacy-appropriate signal or a visitor's own selection, with a complete default page for everyone else.
What to measure
Compare qualified conversions and sales progression within each segment against its default experience. Check that personalization does not confuse or exclude visitors.
Keep in view
Do not infer sensitive traits or expose private deal details. Unknown referrers and missing consent should fall back to the default experience.
Link to this idea
A tierAI search

Agent-readable twin

Publish a clean Markdown version of key pages for AI tools to read. Check weekly whether sampled citations change.

Plan the first test
First test
Create a text-first twin of one important page, preserving its claims, links and canonical HTML reference. Keep both versions in sync and make the Markdown discoverable.
What to measure
Check that the twin is accessible and accurate, then compare citations for a stable question set. Record page changes and other activity that could explain a difference.
Keep in view
A Markdown twin does not guarantee discovery, indexing or citations. It should clarify the same content, not serve different claims to people and AI tools.
Link to this idea
A tierAI search

Screenshot scan of AI answers

Use a vision model to inspect what a buyer actually sees in an AI answer, including who gets the link and whose logo appears.

Plan the first test
First test
Capture a permitted screenshot for a documented buyer question, tool, date and viewport. Ask for visible brand, link and logo observations, then have a person check the image against the extracted record.
What to measure
Track visible placement and prominence in a consistent screenshot sample. Separate what was shown from clicks, visits and inquiries.
Keep in view
Screenshots capture one interface state. Vision models can misread text or logos; verify destinations separately and remove personal information before analysis.
Link to this idea
A tierAI search

Bot log intelligence

Read your server logs to understand which AI crawlers visit and what they request. Use the findings to improve access, structure and internal links.

Plan the first test
First test
Work with the site's owner to review authorized, privacy-filtered logs. Group claimed AI crawler requests by URL and status, and validate identity against the provider's guidance where possible.
What to measure
Track crawl errors, access to priority pages and changes in verified crawler coverage after a scoped fix. Review citations separately.
Keep in view
User-agent strings can be spoofed. A crawler visit is not evidence of training use, indexing, a citation or buyer demand.
Link to this idea

From idea to a useful test

One hypothesis.
One clear next step.

Choose a business problem, name the result you want to observe and set a limit on time, traffic and cost before you start. Keep a person responsible for the decisions.

Explore the strategy playbooks