Using AI to Chat with your Online Business

,

Stop building GA4 reports. Conversational analytics lets digital marketers ask plain-English questions of their data and get answers. Free prompt kit inside.

This is the third presentation in our Pocket Conversion Scientist® series.

Your Website Knows More About Your Prospects Than You Think

You just can’t hear it yet.

Every session, every channel, every dropped form, every product page that converts at 0.8% when it should be at 3%. It’s all in your analytics right now.

The problem isn’t the data. It’s how we’ve been accessing it.

Building reports takes time. And even when you do it, you get numbers. Not answers.

I’ve been doing CRO since 2007. We’ve built scorecards, dashboards, custom Google Sheets that pull from the API. We’re pretty sophisticated about this. But nothing has changed our analysis the way this has.

You can now have a plain-English conversation with your analytics data.

Ask it what’s wrong. Ask it what’s working. Ask it why that traffic channel dropped off in March. It will tell you.

Here’s how we do it, and how you can start.

Why Most AI Website Tools Get This Backwards

The AI tools pitching you right now have this backwards.

They look at your website and hand you a list of things to fix. I’ve tried a number of them. We haven’t adopted any.

Here’s why. They skip your data. They start with your pages, not your analytics. So what they give you is best practices dressed up as personalized recommendations.

That’s a problem.

A page that’s performing well is a dangerous page to change. There are more ways to hurt it than help it. If you’re testing ideas that aren’t grounded in your actual data, you’re guessing.

Conversational analytics starts where we start. With the data.

How to Connect AI to Your Google Analytics

There are two ways to connect a language model to Google Analytics.

The first is an MCP server. MCP stands for Model Context Protocol. It’s essentially an API that’s designed for AI to use. Once connected, your language model can query your analytics directly.

Google provides instructions for setting this up. Allocate about 45 minutes. It involves granting credentials and OAuth setup, which gets a little nervifying if you’re not a developer. My advice: paste Google’s guide into your AI and have it walk you through each step.

MeasureU has also published setup guides for Mac and Windows if you want more hand-holding.

The second option is EverOpt, and it’s what we use.

EverOpt connects via the Google APIs directly, which means fewer hallucinations. Most AI tools read a table and fill in the gaps with pattern matching. EverOpt generates JSON and SQL to pull the actual numbers. That’s a meaningful difference when you’re making decisions based on the output.

It also has memory, a daily or weekly briefing agent, and supports Tag Manager, PostHog, and Convert.com.

The Three Prompts That Get You Started

We’ve put together three prompts that get you started.

Here’s what each one does.

The first is the Business Health Audit.

Tell the AI to act as a CRO and business intelligence leader. Ask it to deliver a plain-language health report on your analytics. Tell it not to change anything yet.

It will figure out what kind of business you are just from the data. Then it will tell you which traffic channels are performing, which aren’t, and drill down to the specific campaigns or products causing problems.

The output is a rich report. Read it in full. Or paste it into your ongoing business AI and let it start building strategy from there.

The second prompt is Sellable Traffic.

Not all of your visitors are trying to buy anything.

Bots, AI scrapers, blog readers, existing customers logging in, internal employees — they inflate your session count and suppress your conversion rate. This prompt identifies and removes them so you’re working with numbers that actually mean something.

In one example, removing invalid traffic moved a site’s conversion rate from 3.2% to 3.6%. Without changing a single page.

Once the AI identifies what to exclude, tell it to remember those exclusions for all future analyses. In EverOpt, it saves that to memory and applies the filter automatically.

The third prompt is the Weekly Brief.

Most digital marketers don’t check their analytics every week. This prompt fixes that without requiring you to go anywhere.

Set it to compare 90-day trends against the last 7 days. Ask it to flag traffic channel changes, conversion rate shifts, form abandonment rates, and 404 errors. In EverOpt, paste the prompt into the briefing feature, set your cadence, and it emails you the report.

One of our client briefs flagged a spike in SMS conversions we never would have caught manually. Now we know what drove it, and we can do more of it.

What to Ask Once You’re Connected

Once you’re connected, the questions you can ask are limited only by what’s in your data.

Build funnels. Graph trends. Diagnose why a channel underperformed last month. Examine any A/B tests running on the site.

You can even ask why a variation won or lost. Paste in screenshots of both and the AI will help you build hypotheses. That “why” question has always been the hard one. It still is. But now you have a place to start.

Tell it to interview you about your business. The more context it has, the sharper the analysis gets.

Go Have a Conversation With Your Data

I was a little reluctant to put this together.

This has become a real competitive advantage for us at Conversion Sciences. Being able to walk into a client conversation with an inventory plan or a traffic diagnosis that most agencies don’t even think to produce. That changes the conversation.

But we’ve decided to give away the good stuff.

You’ve got the setup. You’ve got the prompts.

Go have a conversation with your data.

Get the free prompt kit at conversion.science/ad+lp.

RESOURCES

Conversational Analytics Prompt Kit

Connecting AI to Your Analytics

Google MCP Setup

Measure U Guides

Download the Mac Guide
Download the Windows Guide

EverOpt

Get EverOpt

*Affiliate Link

YouTube Video

Conversion Sciences: https://conversionsciences.com

Pocket Conversion Scientist Series

Using AI to Generate Copy that Converts

Ad + Landing Page Alignment

/by

Your Ads Are Making Promises Your Landing Page Experience Doesn’t Keep

,

There is a fundamental disconnect in our organizations that is showing up in our ad campaigns.

One team creates and manages the ads, Meta ads, Google ads, Bing ads. The landing pages from which the ad team chooses were created by another team, the web team.

The ad team has one priority: increase traffic and demand that generates revenue. The web team has a number of goals, most of which involve publishing pages.

These goals are not aligned.

How bad can things get?

We pulled the ad portfolios of dozens of brands and compared them to the landing pages that serve those ads link. We were surprised that so many brands had serious alignment issues:

  • Ads linking to the home page
  • Ads linking to pages without calls to action
  • Landing pages that served dozens of different ads

Here are some of the things I cover in my presentation, which you can watch on YouTube.

Why does ad spend keep rising while conversion rate stays flat?

Because the ad makes a promise (an offer, a headline, a specific claim) that the landing page doesn’t keep, so no amount of extra spend fixes a ceiling caused by the page itself.

What is the Ad Alignment Report?

An AI-generated audit that pulls a client’s full ad portfolio, matches every ad to the landing page it points to, and flags where the offer, headline, or design don’t line up.

What does a 4.7-to-1 ad-to-landing-page ratio actually mean for a business?

It means nearly five different ads, each with a different promise — different offers, headlines, calls to action — are all dumping traffic onto the same page. It’s difficult to create a single landing page that can support so many ads and still have a high conversion rate.

The larger the ad-to-landing-page ratio, the more likely your ad spend is being wasted on misalignment.

Why do so many ads land on the homepage instead of a dedicated page?

Because building and maintaining a landing page for every ad is more work than most teams can keep up with. When the website has no good alternatives, the homepage is chosen by the ad team.

Unfortunately, the homepage isn’t designed to convert ad traffic. It serves many kinds of visitors and can’t speak to any single ad’s specific promise.

What are “The Ten Alignments”?

The ten factors a landing page should match from its ad:

  • the offer
  • headline
  • call to action
  • trust and proof
  • product or service

These are the five that matter most.

Add to these:

  • image
  • brand
  • tone
  • URL
  • color

These add extra connections for those who click on the ads.

Does every alignment factor matter equally?

The first five (offer, headline, CTA, trust/proof, product) carry most of the weight. Get those wrong and the visitor’s trust breaks before the smaller details even register.

Why do social ads and search ads need different alignment strategies?

Because social ads interrupt someone who wasn’t looking for you, while search ads meet intent someone already had. The landing page has to pick up the conversation differently in each case.

What happens when an ad’s specific offer never appears on the landing page?

The visitor who clicked specifically for that offer, like AVE’s “4 Months Free,” arrives to find no trace of it and has no reason to trust the rest of the page.

Why does a headline mismatch hurt more than it seems like it should?

It removes the visitor’s confirmation that they landed in the right place, so instead of reading on, they’re left wondering if they clicked the wrong link.

Why does tone matter as much as offer or headline?

Because a warm, personal ad voice followed by a clinical form-heavy page creates a jarring shift at the exact moment trust needs reinforcing, and visitors disengage from that whiplash.

Can this alignment work actually be automated?

Yes. AI trained on landing page best practices can generate a matching page for each ad’s specific offer, headline, and tone, which is what the AVE and Careforth “Safe at Home” rebuilds demonstrate.

What do you do once a landing page is already well aligned?

Test a differentiators band (trust signals placed above the fold) or a Good-Better-Best pricing ladder that anchors value with a high first price, both shown on Uline’s pages.

How should a business manage alignment across dozens or hundreds of landing pages?

Either use personalization and A/B testing tools to adjust the page per visit for smaller portfolios, or generate aligned pages with AI and route ad traffic automatically for larger ones.

Why does this matter more now than it used to?

Because Google’s AI and Performance Max now factor landing page quality into ad delivery itself, so a weak page doesn’t just lose conversions anymore, it quietly drags down the ads pointing to it too.

Frequently Asked Questions

What is ad-to-landing-page alignment?

Ad-to-landing-page alignment is how closely a landing page matches the offer, headline, and promise made in the ad that sent the visitor there. When the ad and the page tell different stories, visitors lose the sense that they landed in the right place and leave before converting.

Why do my ads convert but my landing page doesn’t?

Usually because the landing page doesn’t keep the specific promise the ad made. A generic homepage or product grid can’t speak to five different ad offers at once, so most visitors arrive expecting something the page never delivers.

What is message match in PPC advertising?

Message match is the practice of echoing an ad’s exact headline, offer, and value proposition on the landing page it links to. It’s one of the core inputs Google uses to score landing page experience, which directly affects Quality Score and cost per click.

Why doesn’t my ad’s offer show up on my landing page?

This usually happens when ad copy and landing page copy are created separately, often by different teams. Without a shared alignment check across the ad portfolio, an offer like “4 months free” can vanish entirely by the time a visitor lands.

What are the most important elements to align between an ad and a landing page?

The five highest-impact factors are the offer, the headline, the call to action, trust and proof, and the product or service featured. Together these carry most of the alignment score. Image, brand, tone, URL, and color refine it further.

Should every ad have its own landing page?

Not necessarily one-to-one, but a portfolio where dozens of ads all route to a single homepage is a strong signal of misalignment. A useful benchmark is your ad-to-landing-page ratio. Anything close to 5:1 usually means most ads are landing somewhere generic.

How does landing page alignment affect Google Ads Quality Score?

Landing page experience is one of three components of Quality Score, alongside expected click-through rate and ad relevance. A page that doesn’t match its ad’s message can raise cost per click even when the ad itself performs well.

Can AI generate landing pages that automatically match each ad?

Yes. AI trained on landing page best practices can generate a page variant for each ad’s specific offer, headline, and tone, then route that ad’s traffic to its matching page automatically. This is especially useful for portfolios too large to manage by hand.

RESOURCES

YouTube Video

Download the slides

Conversion Sciences: https://conversionsciences.com

Last webinar: Using AI to Generate Copy that Converts

/by

Using AI to Generate Copy that Converts

,

Audio Podcast

Podcast: Play in new window | Download

AI can write website copy that outperforms 80% of what is online today.

You just have to teach it who it is writing for.

In this episode of Intended Consequences, Conversion Sciences founder Brian Massey shows you how to use AI to generate website copy that actually converts.

The secret is not a better prompt.

It is writing for the four ways people make buying decisions.

You will learn the Modes of Research framework, first published in “Waiting for Your Cat to Bark,” and how to map it onto Myers-Briggs types so any language model speaks your language.

Then you will watch live rewrites that turn flat, jargon-filled copy into messaging built for Competitive, Methodical, Spontaneous, and Humanist visitors.

By the end you can build your own AI messaging agent in ChatGPT, Claude, or Gemini and let it do the rewriting for you.

WHAT YOU WILL LEARN

  • Why most B2B copy sounds the same and caps your conversion rate
  • The four research modes and the buyer behind each one
  • How to use Myers-Briggs as a shared vocabulary with any AI
  • The simple prompt that teaches your chatbot to rewrite by mode
  • How to generate personas straight from a URL
  • How to A/B test copy that is finally different enough to win
  • How to build a reusable AI messaging agent for your brand

CHAPTERS

00:00 Why AI copy beats 80% of website copy
01:30 Styrofoam copy and the conversion ceiling
02:40 How our own biases sabotage copywriting
04:10 ICPs and the four-persona problem
05:40 Corner cases: copy big enough to A/B test
06:00 The 4 Modes of Research framework
06:50 Competitive and Methodical buyers
08:00 Spontaneous and Humanist buyers
09:30 Placing copy on the page by buyer mode
10:30 Why language models beat humans at this
11:20 Myers-Briggs as a shared language with AI
14:00 The simple prompt to train your chatbot
15:00 Generating personas from a URL (Calm.com)
17:40 Rewriting copy for each mode, live
24:00 B2B example: HR services, CHRO vs CFO
29:50 Laying out multiple voices on one page
31:00 Q&A: getting your team to trust AI copy
33:20 Building your own AI messaging agent
38:00 What is next: ad and landing page alignment
38:50 Q&A: CTAs, ad frequency, and brand salience

RESOURCES

Download the slides | Get the Cheat Sheet

Messaging skills and full prompts: https://conversion.science/msg-skills

Conversion Sciences: https://conversionsciences.com

Book: “Waiting for Your Cat to Bark” by Bryan and Jeffrey Eisenberg: https://conversci.com/catbark

Roy H. Williams and the Wizard Academy: https://www.wizardacademy.org

Subscribe for more on conversion optimization, AI, and the experiments behind what actually works.

/by

AI: Why Your Website is Now a Training Tool

,

Podcast: Play in new window | Download

In a recent conversation, Brian Massey and Andy Crestodina explored the shifting landscape of digital marketing as Artificial Intelligence (AI) becomes the “third partner” in the relationship between brands and customers. This shift is fundamentally changing the purpose of a website—from a destination for human eyes to a critical knowledge source for AI training.

As AI models like ChatGPT and Gemini increasingly act as “sales reps” or agents for your brand, the strategy for building and maintaining a digital presence must evolve. Here is the breakdown of how to adapt your website for this new era.

The New Role of Your Website

For decades, websites were built for search engines (SEO) and humans (CRO). Today, a third audience has emerged: AI LLMs (Large Language Models).

  • AI as a Sales Rep: AI helps users decide if a brand is a good option by ingesting data from across the web.
  • Knowledge Sources: Your website is your best chance to tell your full story—who you serve, how well you do it, and your specific credentials—because it is the only digital property where you have unlimited space to provide data to AI knowledge sources.
  • The Conversion Trap: While AI traffic may be a small percentage of total visitors today, these users often arrive “further down the funnel” because they have already had long conversations with the AI about their specific goals and context.

Actionable Steps: How to Optimize for AI and Humans

To remain competitive, marketers must “plug holes” in the data AI currently has about their brands.

1. Prioritize Extreme Clarity and Detail

If you were making a page specifically for an AI bot, you would prioritize maximum information over user experience (UX). While you shouldn’t puke “sitemap.xml” onto your homepage, you should ensure your sales and service pages are packed with specific data.

  • List Proof Points as Text: Don’t rely solely on images or trust seals. AI crawlers may not “see” the pixels of an award badge, so you should list these credentials as text to ensure they are ingested into the training data.
  • Define Your “Who” and “How”: Clearly state who you do work for, what the outcomes are, and the credentials of those doing the work.

2. Ditch the Generic Assets

AI learns nothing from generic marketing fluff.

  • Kill the Stock Photos: Brian Massey has long argued against stock photos, but now there is a “moral high ground”: happy people laughing at a laptop tells AI nothing about your brand’s unique value.
  • Stop Trying to Please Everyone: The mistake many marketers make is being too broad. Use your content to disqualify the wrong prospects and better qualify the right ones.

3. Shift Your SEO Strategy

The “crash in click-through rates” for top-of-funnel informational queries means that optimizing articles for generic keywords is less sustainable.

  • Focus on Commercial Intent: Double down on optimizing sales and service pages for traditional search and AI, where visitors still have “visit website intent”.
  • Write for Humans (and Personality): AI can recognize brand personality and humor. Don’t “ring out the cleverness” of your writing; instead, let your brand’s unique voice act as a “culture fit” for potential clients.

4. Feed the AI Beyond Your Website

AI training data isn’t just pulled from your domain; it’s pulled from everywhere.

  • The “Podcast Pitch”: Because Google (which owns YouTube) has access to video transcripts, any podcast or video appearance is a training opportunity. Always give your clear “pitch” during recorded interviews to ensure that language is ingested and associated with your brand.
  • Utilize Other Channels: Since Google is less likely to promote your informational articles today, focus on email, social media, and platforms like LinkedIn or YouTube to reach your audience and train AI models simultaneously.

Finding the “Weakest Link”

Ultimately, marketing should still start from the bottom up. Before worrying about ranking for a new informational keyword, Andy Crestodina recommends looking at the end of the conversion chain.

  • Optimize CTAs: Higher click-through rates on calls to action offer immediate value.
  • Fix Contact Forms: Use data to find and eliminate friction in your final contact steps.

The digital landscape is moving toward a hybrid model where AI handles the “deep research” and humans make the final connection based on trust and brand experience. By treating your website as a rich data source for AI today, you ensure you are the brand recommended tomorrow.

Taking a cue from Andy’s own suggestion during the interview to “give it to a tool that synthesizes or summarizes it for us,” this blog post was generated by a Large Language Model (LLM) using the full transcript of the conversation between Brian Massey and Andy Crestodina.

Checklist: AI-Friendly Website Elements

Use this checklist to ensure your “AI sales reps” have the best information to recommend your brand.

1. Essential Brand Data (The “Who, How, and Why”)

  • Detailed “Who We Serve”: Explicitly list your target industries and customer types in plain text.
  • Specific Outcomes: Describe the exact results you achieve (e.g., “reduced churn by 20%”) rather than using generic marketing fluff.
  • Credentials as Text: Don’t just show award logos; list awards, certifications, and years of experience as crawlable text.
  • Author Bios: Include detailed bios for your team with links to LinkedIn or other publications to establish authority.

2. Conversational & Structural Clarity

  • Direct Answers Top-of-Page: Place the main answer to a query within the first 50–60 words of a section.
  • Extreme Clarity: Abandon “clever” copywriting in favor of high-detail language that defines concepts clearly.
  • TL;DR / Key Takeaway Blocks: Add short summaries at the top or bottom of long sections to make them easy for AI to synthesize.
  • FAQ Sections: Use real customer questions as headings to capture long-tail AI queries.

3. Technical “Machine-Readable” Elements

  • Logical Heading Hierarchy: Use only one H1 per page, followed by properly nested H2 and H3 tags.
  • Schema Markup: Implement JSON-LD schema, specifically Organization, Product, Review, and FAQPage.
  • Descriptive Alt Text: Include context that explains the value of the image (e.g., “Chart showing 40% ROI increase for B2B client”).
  • Semantic HTML Tags: Wrap your content in tags like <main>, <article>, and <section> so AI knows which parts are primary.
  • llms.txt File: Create a simple Markdown file in your root directory to provide a “map” specifically for AI crawlers.

4. Brand Personality Training

  • Consistent Tone: Maintain a unique brand voice across all pages so AI can associate specific traits with your brand.
  • Original Data/Insights: Publish original research or case studies that only your brand can provide, making you a “high-quality” source.

Links

Video Podcast

Orbit Media Studios

Oliver Emberton

/by