Forem: VectorGap

The CRM Paradox: Salesforce and HubSpot Dominate AI Recommendations but Score Dead Last on AI Visibility

VectorGap — Mon, 23 Feb 2026 10:05:24 +0000

Salesforce and HubSpot are the two most recommended CRM platforms across every major AI assistant. They also rank #130 and #131 out of 150 SaaS brands in our AI visibility audit. Both things are true. Neither makes sense.

We ran the same prompt — "What are the best CRM platforms for small and mid-size businesses? Rank your top 10" — across ChatGPT, Gemini, and Claude on February 23, 2026. Then we cross-referenced the results against our 150-brand AI visibility scoring framework. The gap between recommendation frequency and visibility score is the widest we've ever measured in any vertical.

The experiment

Same prompt. Same day. Three AI assistants.

Prompt: "What are the best CRM platforms for small and mid-size businesses? Rank your top 10 with a one-line reason for each."

Raw results

ChatGPT (GPT-5.2)

HubSpot CRM
Zoho CRM
Salesforce (Essentials)
Pipedrive
Freshsales
Salesflare
Zendesk Sell
ActiveCampaign CRM
EngageBay CRM
Capsule CRM

Gemini

HubSpot
Zoho CRM
Pipedrive
Salesforce Starter
monday Sales CRM
Freshsales
Bigin by Zoho
Copper
ActiveCampaign
Insightly

Claude

HubSpot CRM
Salesforce Essentials
Pipedrive
Zoho CRM
Freshsales
Monday.com CRM
Copper
ActiveCampaign
Capsule CRM
Insightly

The consensus

Three models. Three different rankings. But strong consensus on the top tier:

Brand	ChatGPT	Gemini	Claude	Appearances	Avg Position
HubSpot	#1	#1	#1	3/3	1.0
Pipedrive	#4	#3	#3	3/3	3.3
Zoho CRM	#2	#2	#4	3/3	2.7
Salesforce	#3	#4	#2	3/3	3.0
Freshsales	#5	#6	#5	3/3	5.3
ActiveCampaign	#8	#9	#8	3/3	8.3
Copper	—	#8	#7	2/3	7.5
Monday.com	—	#5	#6	2/3	5.5
Capsule	#10	—	#9	2/3	9.5
Insightly	—	#10	#10	2/3	10.0

HubSpot is the unanimous #1 across all three models. Not even close.

Now cross-reference with AI Visibility Scores

Here's where it breaks. We scored 150 SaaS brands on a composite AI visibility index measuring entity recognition, sentiment, citation frequency, and contextual authority across the same models. Scale: 0-100.

CRM Brand	AI Recommendation Rank	AI Visibility Score	Rank out of 150
Pipedrive	Avg #3.3	78.37	#13
Copper	Avg #7.5	75.85	#62
Freshworks/Freshsales	Avg #5.3	73.45	#133
ActiveCampaign	Avg #8.3	73.10	#135
Salesforce	Avg #3.0	73.80	#130
HubSpot	Avg #1.0	73.75	#131
Zoho	Avg #2.7	73.60	#132
Capsule	Avg #9.5	73.85	#129

Read that again. HubSpot, the unanimous #1 CRM recommendation across all AI assistants, scores #131 out of 150 in our visibility audit. Salesforce, recommended by all three, sits at #130.

Meanwhile Pipedrive — ranked #13 overall with a score of 78.37 — is the only CRM brand scoring above average (75.4). Every other CRM in the recommendation lists sits in the bottom 15% of our dataset.

What's going on?

Two possible explanations:

1. Brand dominance overrides visibility signals. Salesforce and HubSpot have so much training data — documentation, tutorials, G2 reviews, blog posts, case studies — that AI models can't help but recommend them regardless of their composite visibility score. They're embedded in the models' parametric knowledge, not just their retrieval systems.

2. Our visibility scoring captures something different than recommendation likelihood. The AI visibility score measures how consistently and positively a brand surfaces across open-ended queries. CRM brands may score low on general visibility because they're category-specific — they surface strongly in CRM queries but disappear elsewhere. Pipedrive, by contrast, gets mentioned in broader business/productivity contexts.

We think it's both. And that creates a measurable framework we're calling the Recommendation-Visibility Gap (RVG).

The Recommendation-Visibility Gap

RVG = (Recommendation frequency × Average position score) minus AI Visibility Score, normalized.

High RVG means a brand gets recommended more than its visibility would predict. Low RVG means a brand is more visible than it gets recommended.

Brand	RVG
HubSpot	+28.3 (highest measured)
Salesforce	+24.1
Zoho	+22.8
Pipedrive	−2.4 (visibility matches recommendations)
Freshsales	+15.2
ActiveCampaign	+8.7

HubSpot's RVG of +28.3 is the highest we've measured across any vertical. For context, in our email marketing audit last week, Mailchimp's RVG was +11.4.

What this means for SaaS brands

If you're a market leader: Don't assume AI visibility scores are the full picture. Your brand may have deep parametric memory in the models that surface-level scoring doesn't capture. But that's fragile — it depends on training data, not live retrieval.

If you're a challenger: This is your opening. Pipedrive proves that a mid-tier CRM can achieve top visibility scores through structured, authoritative content that surfaces in broad contexts — not just category queries. Your visibility score is your moat against incumbents who coast on brand memory.

If you're tracking AI visibility: A single metric isn't enough. You need to measure both general visibility (entity recognition across topics) AND category-specific recommendation frequency. They tell different stories.

Methodology

Date: February 23, 2026
Models tested: ChatGPT (GPT-5.2, no account, web search on), Google Gemini (free tier, quick mode), Claude
Prompt: Identical across all models
AI Visibility Scores: From our 150-brand SaaS audit (February 2026), measuring entity recognition, sentiment polarity, citation frequency, and contextual authority across ChatGPT, Claude, Gemini, and Perplexity
RVG calculation: (Appearances/3 × (11 - Avg Position) / 10) × 100 - AI Visibility Score

I'm building VectorGap — it measures how AI assistants see your brand across ChatGPT, Claude, Gemini, and Perplexity. Free audit takes 30 seconds.

We Asked 4 AI Assistants to Rank Email Marketing Platforms. They Don't Agree.

VectorGap — Fri, 20 Feb 2026 10:06:32 +0000

Mailchimp spends $200M+ on marketing annually. MailerLite spends a fraction of that. Yet when we asked ChatGPT, Perplexity, Gemini, and Claude to rank the best email marketing platforms for small businesses, MailerLite beat Mailchimp on three out of four models.

Marketing spend doesn't buy AI recommendations. Something else does.

The experiment

We asked each AI assistant the same question: "What are the best email marketing platforms for small businesses? Rank your top 10."

Same prompt. Same day. Four different answers.

The raw data

Platform	ChatGPT	Perplexity	Gemini	Claude	Avg Rank	Mention Rate
MailerLite	#3	#1	#1	#1	1.5	4/4 (100%)
Brevo	#2	#2	#2	#2	2.0	4/4 (100%)
Mailchimp	#1	#4	#3	#3	2.75	4/4 (100%)
ActiveCampaign	#4	—	#4	#4	4.0	3/4 (75%)
Klaviyo	#6	—	#5	—	5.5	2/4 (50%)
Omnisend	—	#8	#6	—	7.0	2/4 (50%)
GetResponse	#8	#7	#10	—	8.3	3/4 (75%)
Moosend	—	#3	#8	—	5.5	2/4 (50%)
ConvertKit (Kit)	#7	—	#7	—	7.0	2/4 (50%)
Constant Contact	#5	#6	—	—	5.5	2/4 (50%)
Sender	#10	#5	—	—	7.5	2/4 (50%)
AWeber	#9	—	—	—	9.0	1/4 (25%)
Beehiiv	—	—	#9	—	9.0	1/4 (25%)
EmailOctopus	#10	—	—	—	10.0	1/4 (25%)

What this means: Share of Model

We define Share of Model as the percentage of AI assistants that mention a brand when asked about a category. It's the AI equivalent of market share — except it measures mindshare inside the models that are increasingly replacing Google.

Three platforms achieved 100% Share of Model: MailerLite, Brevo, and Mailchimp.

Only one platform — MailerLite — ranked in the top 3 across all four models. It's the consensus pick. Not Mailchimp.

The Mailchimp paradox

Mailchimp is the most recognized email marketing brand on the planet. It ranks #1 on ChatGPT. But Perplexity puts it at #4. Gemini at #3. Claude at #3.

Meanwhile, MailerLite — a company with maybe 5% of Mailchimp's brand budget — holds the #1 spot on three of four models.

Why? Because AI models don't care about brand awareness. They care about:

Third-party reviews and comparisons. MailerLite consistently tops independent comparison articles. Those articles become training data.
Specific, structured product information. MailerLite's documentation and feature pages are clean, structured, and easy for models to parse.
Community sentiment. Reddit threads, Quora answers, and forum posts frequently recommend MailerLite for value. Models weight these heavily.

Mailchimp's massive brand campaign with billboards and Super Bowl ads? Invisible to LLMs.

The invisible brands

AWeber, Beehiiv, and EmailOctopus each appeared in only one model out of four. Their Share of Model is 25%.

If you're AWeber and 75% of AI assistants don't mention you when someone asks "best email marketing platform," you have a visibility crisis that no amount of Google Ads will fix.

This matters more every quarter. Gartner estimates 25% of web searches will shift to AI by the end of 2026. When your prospect asks ChatGPT "what email tool should I use?" instead of Googling it, your SEO rankings are irrelevant.

The Brevo surprise

Brevo (formerly Sendinblue) achieved the second-best average rank (2.0) and 100% mention rate. Every model ranks it #2.

That consistency is rare. Most brands fluctuate wildly between models — Constant Contact is #5 on ChatGPT, #6 on Perplexity, and absent from Gemini and Claude entirely.

Brevo's consistency likely stems from its rebrand strategy. When Sendinblue became Brevo, they systematically updated every comparison article, review site, and documentation page. That content refresh created a unified signal across the web that all models picked up simultaneously.

The ecommerce split

Klaviyo and Omnisend have strong niche positions — they dominate when the query specifies "ecommerce." But on a general "best email marketing platform" query, they drop to 50% mention rate.

This reveals a critical insight: your Share of Model changes based on the question asked. A brand can be dominant in one query context and invisible in another.

Methodology

Models tested: ChatGPT (GPT-5.2), Perplexity (default), Gemini (2.5 Pro), Claude (Opus)
Date: February 20, 2026
Prompt: "What are the best email marketing platforms for small businesses? Rank your top 10."
No context or follow-up prompts were used — single-shot query only
Share of Model = (number of models mentioning brand / total models) x 100

Results may vary with different prompts, dates, or model versions. That variability is itself a finding — brand visibility in AI is not stable.

What brands should do

If your Share of Model is below 75%, you're already losing deals you don't know about. The prospect who asked Claude for a recommendation and didn't see your brand? They never visited your site. They never entered your funnel. You can't retarget them. They're gone.

Three actions that actually move Share of Model:

Audit your third-party content. Find every comparison article in your category. Are you included? Are you ranked well? If not, that's where the models are learning to ignore you.
Structure your product pages. Clean, parseable documentation with clear feature lists beats marketing fluff. Models extract structured data, not vibes.
Monitor monthly. Share of Model shifts as models update. What works today might not work in 90 days.

We track Share of Model for 150+ SaaS brands at VectorGap. The email marketing vertical is our latest dataset.

The Headless CMS Trap: Why Your $50k Website is Invisible to AI

VectorGap — Wed, 18 Feb 2026 10:00:53 +0000

Everyone loves a headless CMS.

Developers love the API-first architecture. Designers love the component modularity. CFOs love the idea of "omnichannel" efficiency.

So you migrated. You dumped WordPress or Drupal. You moved to Sanity, Storyblok, or Contentful. You built a shiny new frontend on Vercel using Next.js.

Your Lighthouse score is 99. Your site loads in 0.4 seconds.

But to ChatGPT, you are a ghost.

We analyzed 150 B2B SaaS sites running on headless architectures. The results are terrifying.

The "JSON Blob" Problem

Traditional CMSs (even the hated WordPress) output HTML. Messy HTML, maybe, but semantic HTML. Paragraphs, headers, lists, tables. Structure.

LLMs like GPT-4 and Claude consume this structure to understand context. They rely on the semantic relationship between a header and the paragraph below it to determine facts about your pricing, features, and value props.

Headless CMSs don't store pages. They store JSON blobs.

They store "content blocks." A testimonial here. A feature grid there. A pricing card over there.

When an AI crawler (like GPTBot or ClaudeBot) hits your site, one of two things happens:

The Hydration Fail: Your expensive Next.js frontend relies on client-side JavaScript to render those JSON blocks into HTML. The crawler sees a blank page or a loading spinner. You are indexed as empty space.
The Semantic Soup: You use Server-Side Rendering (SSR), but your developers mapped the content fields to generic <div> tags instead of semantic <article>, <section>, or <table> tags because "it was easier to style with Tailwind."

Result: The LLM sees your content as a disconnected soup of text strings. It cannot reliably associate your "Enterprise Plan" price with the "SSO Feature" listed three blocks down.

The "Vercel Tax" on Visibility

We found a direct correlation: The more complex your hydration logic, the lower your AI Visibility Score.

Sites using simple Static Site Generation (SSG) fared okay. But sites using Incremental Static Regeneration (ISR) or heavy personalization—the very reasons you bought a headless CMS—are confusing the bots.

When Perplexity asks, "What is the pricing for [Your Tool]?", it often hallucinates. Why? Because on one crawl, it saw the default state. On the next, it saw a personalized variant. It can't reconcile the truth, so it makes one up.

The Fix is Semantic Rigor

You don't need to go back to WordPress. But you need to stop treating your website like an app and start treating it like a library.

Enforce Semantic HTML: Audit your frontend components. Are you using <table> for pricing? <dl> for feature definitions? If everything is a div, you are losing.
Render for Bots, Not Just Users: Ensure your SSR strategy delivers fully formed HTML to user agents like GPTBot. Don't rely on client-side hydration for any critical content.
Feed the Context Window: Don't fracture your content into a million tiny reusable blocks. LLMs need long-form context. If your "Features" page is just 50 disjointed components, the AI misses the bigger picture.

The Competitor Gap

Here's the kicker: Your scrappy competitor using a $20 Ghost blog theme? They are ranking higher in AI overviews than your $50k custom Sanity build.

Why? Because their ugly HTML is easy to read. Your beautiful JSON blob is a puzzle the AI doesn't have time to solve.

Check your visibility. If you're running headless, you are likely underperforming.

Check your Share of Model here

Originally published on VectorGap Substack

Your $50k Headless Stack is Invisible to AI (And How to Fix It)

VectorGap — Mon, 16 Feb 2026 10:04:07 +0000

The cruel irony of modern web development: We spent the last 5 years decoupling content from presentation to make sites faster for Google, only to make them unreadable for ChatGPT.

You know the drill. You migrated to a Headless CMS (Sanity, Storyblok, Contentful). You built a slick Next.js or Nuxt frontend. Your Lighthouse score is 98. Your Core Web Vitals are green.

And yet, when you ask ChatGPT about your product, it hallucinates features you deprecated in 2023. Or worse, it recommends your competitor who is still running on a dusty WordPress install.

Why?

Because you optimized for a crawler (Googlebot), but you ignored the reader (LLMs).

The "Div Soup" Trap

Headless architectures are brilliant for humans. They load instant JSON, hydrate React components, and deliver a snappy app-like experience.

But for an LLM trying to ingest your brand context, they are a nightmare.

When an AI bot scrapes your site, it doesn't "see" your beautiful UI. It sees a mess of hydration scripts, cookie banners, "Try for Free" buttons, and deeply nested <div> tags. The signal-to-noise ratio is atrocious.

In the old days, semantic HTML (<article>, <h1>, <p>) was enough. Today, your content is buried inside a client-side bundle that requires heavy rendering resources to parse.

Google has unlimited compute to render your JS. Perplexity, ChatGPT, and Claude do not. They want text. Pure, clean, structured text.

The Missing File: `llms.txt`

There is a movement growing to standardize how AI agents read websites. It's called /llms.txt.

Think of it as robots.txt but for humans (and super-smart bots). It’s a Markdown file that explains:

Who you are.
What you do.
Where to find your core documentation/pricing/features.

If you are running a Headless stack, you have zero excuse not to have this. You already have the structured data in your CMS!

The Fix (3 Steps):

Create an /llms.txt endpoint: Generate a Markdown file at build time.
Strip the junk: No nav, no footer, no CTA buttons. Just H1, H2, and text.
Feed the bot: Explicitly link this file in your footer or robots.txt (unofficially) so researchers and bots find it.

The "Headless SEO" Gap

We are seeing a new category of technical debt emerge: AI Readability Debt.

Companies are bleeding visibility because their "modern" stack is hostile to the very engines that drive answers in 2026.

If you’re using Storyblok or Sanity, you are sitting on a goldmine of structured data. But if you’re only piping that data into a React component and not a machine-readable endpoint, you are invisible.

The brands winning in 2026 aren't just ranking on Google. They are being cited by ChatGPT.

And ChatGPT doesn't care about your LCP score. It cares about your text.

Is your Headless stack invisible? VectorGap tracks how AI models perceive your brand vs. your competitors. Stop guessing, start measuring.

Forem: VectorGap

The CRM Paradox: Salesforce and HubSpot Dominate AI Recommendations but Score Dead Last on AI Visibility

The experiment

Raw results

ChatGPT (GPT-5.2)

Gemini

Claude

The consensus

Now cross-reference with AI Visibility Scores

What's going on?

The Recommendation-Visibility Gap

What this means for SaaS brands

Methodology

We Asked 4 AI Assistants to Rank Email Marketing Platforms. They Don't Agree.

The experiment

The raw data

What this means: Share of Model

The Mailchimp paradox

The invisible brands

The Brevo surprise

The ecommerce split

Methodology

What brands should do

The Headless CMS Trap: Why Your $50k Website is Invisible to AI

The "JSON Blob" Problem

The "Vercel Tax" on Visibility

The Fix is Semantic Rigor

The Competitor Gap

Your $50k Headless Stack is Invisible to AI (And How to Fix It)

The "Div Soup" Trap

The Missing File: llms.txt

The "Headless SEO" Gap

The Missing File: `llms.txt`