Abraham Quiros Villalba AI Tool
I Almost Shut Down My Newsletter Last Year — Then AI Reader Engagement Tools Saved It
Back in late 2025, I was staring at my Substack dashboard feeling pretty defeated.
Open rates had dropped from 42% to 19% in about eight months. Comments had basically stopped. People were subscribing and then just… ghosting. I’d spend six hours writing something I was genuinely proud of, hit publish, and hear crickets.
I run a mid-sized tech and productivity newsletter — nothing huge, around 14,000 subscribers at the time — and I was ready to just fold it into a “monthly digest” and call it a day.
Then a friend who runs a much bigger publication (she’s at around 90,000 subscribers now) told me she’d rebuilt her whole engagement strategy around AI tools in 2026. Not AI writing the content — she was clear about that — but AI helping readers actually connect with content that was already good. Abraham Quiros Villalba AI Tool
I was skeptical. I’d tried “AI personalization” tools before and they felt gimmicky, like a chatbot popup nobody asked for. But I gave it a real shot for about seven months, and I want to walk you through what actually worked, what wasted my time, and what I’d tell anyone starting from scratch right now.
The Real Problem Wasn’t My Writing
This was the first thing I had to accept, and it stung a bit.
I kept assuming my drop in engagement meant my writing had gotten worse or boring. So I rewrote headlines, tried punchier intros, all of that. Barely moved the needle.
Turns out the actual issue was timing and relevance, not quality. My readers were a mixed bag — some wanted deep technical breakdowns, some just wanted quick news, some only cared about AI tools, others only cared about productivity apps. I was sending everyone the exact same email at the exact same time.
That’s basically the core problem AI reader engagement tools in 2026 are built to solve: matching content to the right person, at the right moment, without you manually segmenting a spreadsheet for hours. Abraham Quiros Villalba AI Tool
What I Actually Started Using
I’m not going to pretend every tool I tried was great. Here’s the honest breakdown.
1. Smart segmentation inside my existing platform
I use beehiiv now (switched from Substack in early 2026 partly for this reason), and its AI-driven segmentation was the first real win. It looks at what people click, how long they stay on a post, and what topics they engage with, then groups them automatically.
I didn’t have to build these segments myself..
2. AI-assisted subject line testing
I started running my subject lines through Claude before sending, just asking it to give me three alternate versions with different angles — curiosity-driven, benefit-driven, and blunt/direct. Then I’d A/B test the top two using beehiiv’s built-in testing.
My open rates climbed from 19% back up to around 34% over four months. Not overnight magic, just steady improvement from testing instead of guessing.
3. Chat-based content assistants on the site itself
This one surprised me the most. I added a simple embedded chat widget (I used one built on top of a basic RAG setup, nothing fancy) so readers could ask questions about older posts instead of scrolling through my archive.
I genuinely thought nobody would use it. About 11% of my active readers used it in the first month, mostly asking things like “what did you recommend for note-taking apps last year” or “summarize your take on X tool.”
4. AI-generated content summaries at the top of long posts
For my longer pieces (2,000+ words), I started adding a short “quick take” summary generated with AI assistance, then edited by me for accuracy. Readers who skimmed the summary were noticeably more likely to click into the full article than readers who saw nothing. Abraham Quiros Villalba AI Tool
Step-by-Step: How I’d Set This Up If Starting Today
If you’re where I was — decent content, sinking engagement — here’s the order I’d actually do things in.
Step 1: Audit your actual engagement data first Don’t touch any AI tool yet. Look at your last 20 sends. What topics got clicks? What got ignored? You need a baseline or you won’t know if anything’s working.
Step 2: Pick a platform with built-in AI segmentation Don’t bolt on five separate tools. Beehiiv, ConvertKit, and Ghost (with some plugin help) all have decent native options now.
Step 3: Use AI for subject lines and summaries, not full drafts This is where I’ve seen the best return. Readers can smell fully AI-written content from a mile away, and it kills trust fast. Use AI to sharpen what you’ve already written, not replace it.
Step 4: Add one interactive element, not ten I made the mistake of trying to add a chatbot, a quiz, a poll, and a recommendation engine all in the same month. It was overwhelming for readers and for me. Pick one. See if it sticks. Then add another.
Step 5: Review performance monthly, not daily Checking analytics every day made me anxious and led to knee-jerk changes. Monthly reviews gave me enough data to actually see trends instead of noise.
Mistakes I Made
I over-personalized too fast. I tried sending five different versions of the same newsletter to five segments in one week. It was a nightmare to manage and honestly confused some readers who noticed the inconsistency when they compared notes in comments.
I trusted AI-generated engagement metrics without double-checking. One tool told me a huge chunk of my readers were “highly engaged” based on scroll depth alone. Turns out a bunch of that was people leaving tabs open, not actually reading. Don’t take one metric as gospel.
I ignored the readers who didn’t want AI-anything. A decent chunk of my long-time subscribers pushed back hard on the chat widget and the AI summaries. I added a simple toggle to turn off the “quick take” summaries for people who just wanted the raw article. That single change reduced complaints significantly.
I forgot to tell people what was AI-assisted. Transparency mattered more than I expected. Once I added a small note saying “summary drafted with AI assistance, reviewed by me,” trust actually went up, not down.
Real Examples From Other Publishers
A friend running a food blog told me she uses AI-driven “related content” recommendations at the end of posts, and it’s bumped her average session time by almost two minutes per visitor. Not huge, but compounding over months, that’s a lot more ad impressions and better SEO signals.
Another publisher I know in the finance newsletter space uses AI to flag which past articles are getting a sudden traffic spike (usually because of breaking news), then automatically surfaces those in his weekly send. That’s saved him from missing timely opportunities more than once. Abraham Quiros Villalba AI Tool
None of these are flashy. They’re small, boring, practical wins that add up.
Tools Worth Actually Looking At
If you want a starting list instead of researching from scratch:
- beehiiv — built-in AI segmentation and recommendation blocks
- Ghost — solid for self-hosted publishers, pairs well with third-party AI plugins
- Claude or ChatGPT — for subject line testing, summary drafting, and editing help (not full ghostwriting)
- Frase or MarketMuse — helpful for understanding what topics your audience actually searches for
- Simple embedded chat widgets — for letting readers “talk” to your archive instead of searching manually
Final Thoughts
Honestly, none of this fixed my newsletter overnight, and I’m suspicious of anyone who tells you it did for them. What it did was give me better information and remove some of the manual grunt work so I could focus on actually writing good stuff.
The publications doing well with AI reader engagement right now aren’t the ones automating everything. They’re the ones using it to understand their readers a little better and get out of their own way a little more.
If you’re feeling stuck like I was, start small. Pick one tool. Give it two months. Look at the actual numbers before deciding if it’s working. That’s really it.
Leave a Reply