Zero Traffic for 3 Months, Then 8,500 Daily Impressions in 2 Weeks: My SEO + GEO Playbook

Votrix AI — Sat, 24 Oct 2026 14:08:57 +0000

Look at the chart first. This is a new site I only started working on properly in October. The domain had been sitting there for a while, but from July to September nobody touched it, and the Search Console line barely left the floor: a few clicks and a few dozen impressions a day.

In early October I rebuilt it with the process below. Two weeks later, daily impressions went from a few dozen to 8,546, and daily clicks went from single digits to the twenties and thirties.

I've done SEO for a lot of businesses over the past few years, and nothing here is a new trick. It comes down to one thing: rethinking every page from the reader's point of view. Here's the full breakdown.

Why most sites never take off: the 4 problems I see most

Almost every site I've taken over had the same problems. This one did too.

1. The blog talks about the product instead of solving the reader's problem

Someone searching "how to build a website for a small shop" wants to learn how to do it. But most company blogs are full of "our features" and "why we're the best". Readers land, learn nothing, and leave within seconds. Google and AI assistants can both tell, and neither will recommend that page.

2. Comparison pages have no real point of difference, and there are too few of them

Most sites have one or two "us vs. competitor" pages that say "we're easier and cheaper". After reading, the visitor still doesn't know: what's actually different, and which one should I pick for my situation?

3. Pages aren't written for GEO

GEO (Generative Engine Optimization) means getting ChatGPT, Perplexity and other AI assistants to mention and cite you in their answers. An AI cites a page when it can lift an answer straight out of it. If the key information is buried in long paragraphs, or the page takes three paragraphs to get to the point, the AI skips it and cites someone else.

4. Too few pages, too few keywords, English only

A site with a dozen pages, each targeting one or two keywords, simply has too few ways to be found. Staying English-only also leaves smaller-language markets on the table, where competition is far lower.

My process

Step 1: Pick keywords with an SEO tool, not by gut feeling

Use Search Console, Ahrefs, Semrush, DataForSEO or similar, and look for three things:

Low difficulty. A new site chasing big head terms is wasting its time.
Clear intent. Searches like "X alternatives", "X vs Y" and "best X for small teams" come from people who are already choosing a product.
Weak current results. If the pages ranking now are mediocre, you can beat them.

AI is great for brainstorming more phrasings, but always check search volume in a real tool. The volumes AI gives you are mostly made up.

Step 2: Before writing, answer one question: what will the reader learn from this page?

This is the most important step. Before writing any page, finish this sentence: "After reading this page, the reader can ____."

❌ "learn about our product's features"
✅ "tell who Semrush and Ahrefs are each best for, how much they cost, and which one to choose"

If you can't finish the sentence, don't write the page. If you can, build the whole page around it, and only bring up your own product where it genuinely helps.

Step 3: Build lots of comparison pages, each with a clear point of difference

Comparison pages are the best return on effort in both SEO and GEO, because people searching comparisons are close to buying. For this site I built 114 comparison pages in three types:

X vs Y, like Semrush vs Ahrefs or Lovable vs Bolt.
X alternatives, like Lovable alternatives or Semrush alternatives.
Best X, like best AI SEO tools or best AI marketing tools.

Every page has to cover four things:

A one-line verdict: who A is for, and who B is for.
The real differences: only the 3 to 5 points that actually differ, not 30 rows of "both support this".
Pricing, with actual numbers.
The competitor's strengths. A comparison that only praises one side isn't trusted by readers, or by AI.

A weak comparison page vs. a strong one:

	Weak	Strong
Opening	"We're the industry leader in…"	"Pick A if you're on a budget and need to start fast; pick B if you need deep data."
Comparison	30-row feature list, mostly shared	4 real differences, each with its impact explained
Pricing	"Affordable pricing"	"A is $19/month, B is $49/month, and B adds XX"
Competitor	Only weaknesses	When the competitor is the better choice
Ending	"Sign up now"	"If you're in situation XX, choose XX"

Why comparison pages work so well for GEO: a huge share of the questions people ask AI are comparisons. "Is A or B better?" "What's a cheaper alternative to X?" To answer those, an AI needs pages with a clear structure, a comparison table and a definite verdict.

Step 4: Write for GEO

Answer in the first paragraph. No warm-up.
Use the reader's question as the heading, like "Is Semrush or Ahrefs better for a small team?"
Use tables, steps and specific numbers. AI won't quote "significantly improves".
Say what it is, who it's for and what it costs. These are what an AI looks for when recommending a product. If it can't find them, it skips you.
Make sure AI can read the page. Serve real HTML, and don't block GPTBot, ClaudeBot, PerplexityBot and other AI crawlers in robots.txt. I also added an llms.txt file at the site root listing the key pages.

Step 5: Scale up volume and languages, without dropping quality

Only once every page holds up on its own, start scaling. The site now has 514 pages:

Spread across keywords. Break each big topic into specific questions, one page per question. Besides the comparison pages, there are 20 guides, 23 blog posts and 28 solution pages written for specific industries and use cases.
Add smaller languages. I built 124 pages each in Spanish and Chinese, plus a dozen or so core pages each in German, French, Japanese, Korean and Portuguese. The same keyword is often far less competitive outside English.

The quality bar: no machine translation dumps. Re-pick keywords based on how locals actually search, and swap in local examples and pricing. Better to skip a language than to publish weak pages, because low-quality pages drag down the whole site.

Step 6: Review every month

Track just two sets of numbers:

SEO: impressions, clicks and CTR in Search Console, plus rankings for your target keywords.
GEO: pick 20 to 30 questions your customers really ask AI. Every month, ask ChatGPT, Perplexity and Gemini, and note whether you're mentioned, who is, and which pages get cited.

Questions where you're not mentioned become next month's pages. Sites the AI keeps citing are the places to get listed on and mentioned by.

A checklist you can follow today

[ ] Use an SEO tool to pick 30 low-difficulty, clear-intent keywords
[ ] Before writing each page, finish "After reading this, the reader can ____"
[ ] Build 10 comparison pages: X vs Y, X alternatives, Best X
[ ] Give every comparison page a verdict, the real differences, pricing, and the competitor's strengths
[ ] Answer in the first paragraph and use the reader's questions as headings
[ ] Check that robots.txt doesn't block AI crawlers, and add an llms.txt
[ ] Pick 1 or 2 smaller languages and localize properly
[ ] Set 20 benchmark questions and test which brands AI recommends every month

Final thoughts

It's only been two weeks and the line is still climbing. I'll post an update with fresh numbers in a month or two.

After all these projects, the hard part isn't knowing what to do. It's doing it every month: picking keywords, writing pages, building comparisons, testing AI answers, reviewing. So I built this process into my current product, Votrix AI, where AI marketing employees run it automatically. But even without any tool, following the checklist above by hand works.

What's tripped you up with SEO or GEO? I'd love to hear about it in the comments.

DEV Community: Votrix AI