Building in public · Measured every Sunday
We grade the whole local market. Then ourselves against it.
Every week our own check scores this site and nearby competitors on the same rules. Here is what we changed, what moved, and what still has not worked.
Same waters.
Better websites.
The line is flat.
The work isn't.
Our readiness scores are holding steady, which says the foundations are in place. It does not say customers are finding us.
AI readiness
POINTS / 100The homepage's largest paint on a slow phone. Measured after the latest reading, so no score change is claimed for it yet.
Read the chart data
| Reading | Google, us | AI, us | Google, market | AI, market |
|---|---|---|---|---|
| 2026-09-06 | 72 | 77 | 37 | 25 |
| 2026-09-09 | 72 | 77 | 33 | 25 |
| 2026-09-13 | 72 | 77 | 34 | 25 |
Readings before September 2 used different weights and are not comparable, so they are not plotted. The change is recorded in our scoring notes.
What we tried.
What actually moved.
The useful part is the gap between what we expected and what the next reading said.
Rebuilt the site
Clarity and speed
Google readiness stayed at 72. The homepage's largest paint on a slow phone went from 5.3s to 2.95s.
Measured our own ranking
Read the entry →To know where we actually stand
No map results on any of our six searches, one blue link at position six.
Completed our profile links
Make the business easier to identify
AI readiness rose from 67 to 77. No sign yet of more customers.
Kept blaming the Google category
Read the entry →We had already fixed it
Six weeks on the list. Nothing moved, because there was nothing left to correct.
Fixed our Google category
Better rankings
No ranking improvement recorded since.
Put the scoring on a schedule
Read the entry →Numbers nobody can massage
The reading has run every Sunday since, without a miss.
Published an llms.txt
Read the entry →Help AI tools understand the business
Added 20 points to the AI readiness score. No Google change established, and Google says it does not use the file.
Stopped our host scrambling our email
Read the entry →Let assistants read how to reach us
AI readiness contact points went to 10 of 10.
A better score isn't a customer.
This is the part we have to earn next. We keep the gaps in public beside the improvements.
Missing from all six map results.
We ran the six searches our own business should show up for. In the map results we appear for none of them. In the ordinary blue links we appear once, at position six. Read the measurement.
24 pages Google had never fetched.
Found September 17. Our own www address served a duplicate of the site until September 11, and Google spent its visits there instead. The duplicate is fixed and the pages are queued for a fresh look.
0 of 25 ranking points.
Ranking is worth 25 of the 100 Google readiness points. We score 0. The other 75 points say we are ready to rank, not that we do.
A checkpoint.
Not a victory lap.
Measured October 18, 2026, automatically, and published whatever it said.
Google readiness
How ready your site is for Google to find and understand it: pages in the sitemap (20 of 20), business details Google can read (20 of 20), titles and descriptions (15 of 15), page weight and speed (7 of 15), crawl basics (15 of 15), address and area (15 of 15). Those 92 of 100 count for three quarters of the score, then live ranking adds up to 25. 92 x 0.75 + 0 = 69.
AI readiness
How ready your business is to be named in an AI answer: llms.txt file (20 of 20), business details AI can read (20 of 20), answers to buyer questions (12 of 20), a plain email and phone (10 of 10), profile links (15 of 15), and whether assistants actually name us (0 of 15). 20 + 20 + 12 + 10 + 15 + 0 = 77.
Worth knowing: 20 of those AI points come from an llms.txt file. Google says its search does not use these files today. We keep one because it costs nothing to maintain and some AI tools already read it, so the points are readiness, not proof anyone is reading it.
How the number is made.
Fetch the pages.
A small program on a weekly schedule reads this site and checks what search engines can see.
Use the same rules.
Our site and 6 nearby competitors are scored against the same list.
Save the reading.
Every Sunday the result is written to a table. Nobody types the number in by hand.
Publish it.
Up or down, it goes on this page, and the changes we made go in the notes below.
The notes behind the work.
Each entry is a dated change to this site and what happened next.
Charted / Method
What a directory listing is worth to a local business
We list Narrows Web Co. on Google, Bing, Apple Maps, Yelp, Clutch and more. Here is what each kind of listing does for you, which ones link back, and why we traded a link for a spot on DesignRush.
Read post →Charted / Method
One site, one address
If your site answers at both www and the bare domain, search engines can index two copies and crawl each one less. Here is the ten second check.
Read post →Charted / Method
The order a page should be in
Price early. Proof before the ask. One closing step. Here is the order we build a trades homepage in, and the reason for each block.
Read post →Charted / Method
A page per service. A page per town. Price on the homepage.
Every service gets its own page. Every town gets its own page. The price is on the homepage and our client work is linked from it, so you can go and look instead of taking our word.
Read post →Charted / Method
The one Google setting that decides which searches you can win
Your primary category tells Google which searches you are even eligible for. It takes a minute to check. Here is how, and what ours said.
Read post →Charted / Method
Reviews move the needle this year. Here is how we build them.
We measure our own search results the way we measure a client's: run the searches, write down what is there, then work the part that can change this year.
Read post →Horizon / Automation
An agent that grades us weekly
The scoreboard on this page isn't hand-typed. An agent measures narrowsweb.com every Sunday and writes down the number, whatever it is.
Read post →Horizon / GEO
A cheat sheet for the assistants: llms.txt
Most sites make AI guess what they do. We wrote it down in a plain file AI tools can read, so we're ready when they start using it.
Read post →Horizon / Method
Make sure AI can read how to contact you
Plenty of hosts scramble the email address on your site to stop spambots. It also hides you from the assistants people now ask for a recommendation.
Read post →Strategy / Log
Why we publish the whole playbook
Giving away our method doesn't help the competition. Here's the actual math on why.
Read post →Charted / Horizon
What a typical trades site scores
Before you can show a climb you have to know the floor. We scored six nearby sites on the same rubric we use on ourselves. Here is what they got.
Read post →Want your own two scores? Free.
We will run the same check on your business and send it back within one business day. No call required.