Eduard Albu

Posted on Dec 31

Weeks 7-8: Browser-Based AI, Holiday Dips, and Why Third-Party SEO Tools Are Nearly Useless

#buildinpublic #seo #webdev #ai

This is a double update because Christmas and New Year got in the way of shipping.

But even with holiday slowdowns, these two weeks brought two major tool launches, confirmed something I suspected about SEO tools, and revealed just how international QuickTools.one's audience has become.

Here is everything that happened between December 19 and January 1.

The Numbers That Tell a Story

Let me start with the traffic because it shows something interesting about seasonality.

Metric	Value	Change from Week 6
Visitors	1,080	-25%
Visits	1,310	-21%
Page Views	4,790	-20%
Bounce Rate	15%	-3 points
Average Visit Duration	4 min 12 sec	+4%

A 25% traffic drop during Christmas week. This is normal.

People are not searching for PDF tools while opening presents. The interesting part is what happened after the dip.

Traffic recovered to near-normal levels by December 30. Bounce rate improved. Session duration increased. These are signs of a healthy product.

The drop was seasonal, not structural.

Where the Traffic Comes From

The traffic source breakdown continues to surprise me.

Source	Visitors	Share
ChatGPT	810	75%
Google	135	13%
Bing	54	5%
Egypt (direct)	162	15%
Saudi Arabia (direct)	119	11%

ChatGPT still dominates. It sent 75% of referral traffic across both weeks.

But the geographic breakdown is what caught my attention this time.

26% of my traffic came from Egypt and Saudi Arabia. Combined.

I built QuickTools.one in English, for an international audience, without targeting any specific region. And somehow, a quarter of my users are coming from two countries in the Middle East.

I do not know why yet. Maybe these markets are underserved for free PDF tools. Maybe privacy concerns are higher there. Maybe ChatGPT is recommending QuickTools more aggressively in those regions.

Whatever the reason, this is worth watching. If this trend continues, I might need to consider localization or Arabic language support.

Two New Tools Launched

Even during the holidays, I shipped two tools. Both required weeks of research and testing, but they are finally live.

Remove Background (Browser-Based AI)

Remove image backgrounds using AI that runs entirely in your browser. No uploads, no API calls, no privacy concerns.

Try it: https://quicktools.one/remove-background

This tool uses ONNX Runtime Web and a quantized version of the RMBG-1.4 model to run background removal locally. The entire ML model loads in the browser, processes images on your device, and never sends data anywhere.

Technically, this was the hardest tool I have built so far. The model is 176MB, which required careful optimization to avoid killing mobile browsers. I added progressive loading, WebAssembly acceleration, and fallback paths for unsupported devices.

The keyword research justified the effort:

"remove background from image" gets 201,000 monthly searches with difficulty 61
"remove image background" gets 135,000 searches with difficulty 58
"background remover" gets 110,000 searches with difficulty 65

These are competitive keywords. But the privacy-first positioning gives QuickTools.one a real differentiator. Most background removal tools upload your images to servers. Ours does not.

For photographers working with client photos, designers handling confidential mockups, or anyone who just values privacy, this matters.

Convert to WebP

Convert images to WebP format for faster loading and smaller file sizes. Supports batch conversion and quality control.

Try it: https://quicktools.one/convert-to-webp

This one targets web developers optimizing site performance. WebP images are 25-35% smaller than JPEGs at the same quality level, which directly impacts Core Web Vitals scores.

Keyword research revealed a golden opportunity here:

"webp convert png" gets 33,100 searches with difficulty 17
"png to webp" gets 27,100 searches with difficulty 63

Same search intent. Massive difference in difficulty.

I optimized the page around "webp convert png" and related variations like "convert image to webp" (22,200 volume, difficulty 18). Much easier to rank than targeting "png to webp" directly.

With these two tools, QuickTools.one now has 17 live tools.

The Third-Party SEO Tool Problem

This week I discovered something that completely changed how I evaluate backlinks.

I was checking Ubersuggest to see how many referring domains QuickTools.one had. It showed 6 backlinks.

Then I checked Google Search Console. It showed 180.

That is a 30x difference between what third-party tools show and reality.

And it gets worse. Earlier in the project, Ubersuggest showed 4 backlinks when GSC showed 56. The gap has been consistent throughout.

Third-party SEO tools like Ubersuggest, Ahrefs (free version), and Semrush crawl the web with their own bots. They have limitations:

Their crawlers are slower and less comprehensive than Google's
They prioritize established, high-authority domains
New domains with few backlinks get deprioritized
They simply miss most links to new sites

For a domain less than 8 weeks old, these tools are nearly useless. They will show you maybe 5-10% of your actual backlinks.

Google Search Console shows you what Google actually sees. It is the only reliable source of truth for new domains.

The lesson: do not get discouraged by low backlink counts in third-party tools. Check GSC. The reality is probably much better.

Google Search Console Progress

Speaking of GSC, here are some meaningful numbers:

Average CTR: 24.7%
Average Position: 34.5
Backlinks: 180 (up from 56 in Week 5, +220% growth)

The 24.7% CTR seems impossibly high. Industry average is 2-3%.

But here is what is happening: branded searches.

When someone searches "quicktools.one" directly and I rank position 1, the CTR approaches 100%. These branded queries inflate the overall average.

The average position of 34.5 tells the real story. For non-branded keywords, I am ranking somewhere on page 3-4. That is expected for a domain this young with this domain authority.

The "free plugins equal free backlinks" strategy continues working. Browser extensions and npm packages generate automatic links from monitoring sites.

Supporting Content Published

I published two blog articles during this period:

"How to Remove Background from Image: 6 Free Methods" targets comparison searches like "how to remove image background in photoshop" and "remove image background gimp". The format covers QuickTools, Photoshop, GIMP, Canva, PowerPoint, and Google Slides.

"How to Convert PNG to WebP: 5 Free Methods" targets platform-specific searches like "convert png to webp mac" (difficulty 6) and "convert png to webp photoshop" (difficulty 6). The article covers methods for QuickTools, Mac Terminal, Windows, Photoshop, and Canva.

The "X Free Methods" format works consistently. It captures users who are looking for alternatives, positions QuickTools as one option among many (which builds trust), and targets long-tail keywords that the main tool page cannot rank for.

Directory Submissions This Period

I continued the directory submission grind:

SideProjectors (submitted as showcase)
StartupRanking
StartupTile
PitchWall.co (from Week 6, now live)

The goal is consistent progress. A few submissions per week, tracked in a spreadsheet. None of these will move the needle individually. But over months, they compound into meaningful domain authority.

Week 7-8 Summary

Metric	Value
Tools live	17
New tools	2 (Remove Background, Convert to WebP)
Blog articles	2 new
Visitors (2 weeks)	1,080
Top traffic source	ChatGPT (75%)
Backlinks (GSC)	180 (+220% from Week 5)
GSC average CTR	24.7%
GSC average position	34.5
Tool usage	PDF 91%, Image 9%
Total tool conversions	Around 2,000 operations

Key Lessons from Weeks 7-8

1. Browser-based AI is a real differentiator

Running ML models in the browser is not just a technical curiosity. It solves a genuine privacy concern. For sensitive photos or confidential documents, users appreciate knowing their files never leave their device.

2. Third-party SEO tools fail on new domains

Do not trust Ubersuggest, Ahrefs free, or similar tools for backlink counts. They miss 90%+ of your actual links. Use Google Search Console for truth.

3. Holiday dips are normal

A 25% traffic drop during Christmas week is expected. Do not panic. The underlying trends matter more than weekly fluctuations.

4. International audiences are real

26% of my traffic comes from Egypt and Saudi Arabia. These markets might be underserved. Worth considering for future keyword targeting and possibly localization.

5. Golden keywords hide in plain sight

"Webp convert png" and "png to webp" have nearly identical intent. One has difficulty 17, the other difficulty 63. Always check variations.

6. PDF tools are the core product

91% of tool usage is PDF-related. Users come for PDF work. Image tools are secondary. This should inform future development priorities.