The Live Pilot — Boost Runs Its Own Playbook on Itself

We scored our own website with the exact framework we use for clients, and published the number: 30 out of 100 in July. Since then we have been making the same fixes we prescribe to clients, re-auditing, and publishing every score — including the points we chose to give up, and the day we measured late. An independent second grader scores every rerun, and where the two disagree, the lower number is the one you see here. Want the same done for your business? Contact partnership@boostsystems.ai.

Jul 20 · Baseline

→

Oct 6 am · Summer work

→

Oct 6 pm · Technical package

Where the score stands Verified · October 6, 2026

Each category as of this afternoon, with July's number underneath. The off-page category is the one that has barely moved — and the one that decides most of what comes next.

Readability

6 / 10 +3

was 3 / 10 in July

Substance

3 / 8 -2

was 5 / 8 in July

Trust

3 / 8 +2

was 1 / 8 in July

Presence (off the website)

1 / 14 +1

was 0 / 12 in July

Foundation (technical)

8 / 10 +2

was 6 / 10 in July

Score trajectory

Solid points are verified audits. The dashed line is what the next steps are expected to add; each one is replaced by a real re-audit the day it ships.

What changed, and what each change was worth

Every row is a real change to boostsystems.ai or to Boost's presence elsewhere, with the verified score after it. Red rows are points Boost chose to give up.

Jul 20

Baseline, published

The first audit of boostsystems.ai, scored exactly as a client would be. Nothing off the website existed: no Google profile, no listings, no reviews.

Jul 24

Privacy first: address, phone numbers and founder names taken off the site

A deliberate choice that cost points in two categories — operational detail and a named person behind the work. We would rather show the trade than hide it.

—

-3

Aug 18

A client's result, in a sentence

The Native Tongues case study went live with its score and a client quote in plain text — the first proof on the site an engine could quote.

—

Sep 14

Answering real questions, and saying who we are

A twelve-question FAQ written the way people ask, each answer with a specific fact, plus an About page with the business's vital facts restated in the language machines read natively.

—

Sep 14

Boost goes on the map

A Google Business Profile for Boost, verified and now visible in Google's own data feed — the local layer every assistant checks. Zero reviews yet, which is the next job.

—

Sep 15

Top plan moved to “contact us for pricing”

Fair for done-for-you work, but one of three offers without a price reads as partial transparency. The independent grader marked it down; the lower score stands.

-1

Oct 6

A map for the crawlers, a front door with a name on it

A sitemap with dates, a robots file that welcomes every AI crawler and keeps the signed-in portal out, a plain-text summary for AI agents, a canonical address on every page, and titles that finally say what Boost is and where. Scored 46 by us, 44 by the independent grader; 42 stands.

What comes next Projected

Almost every remaining point sits off the website. Each step below gets its own re-audit the day it lands.

Reviews from the businesses we have worked with

Zero reviews anywhere is the single biggest hole. The first handful of Google reviews, collected and published through our own review feed.

Be listed where ChatGPT looks

The listing service that quietly feeds ChatGPT's local recommendations, plus a Yelp page, under the exact same name, phone and hours as Google.

Someone other than Boost mentions Boost

Today a search for our name belongs to a helicopter-equipment company in North Vancouver. One LinkedIn post a month and one earned mention start to change that.

Cite sources, name a founder, publish numbers about Boost

The site makes claims about the market without linking to a source, names nobody, and never says how many businesses it has audited. All three are a morning's work.

By category: July, this morning, this afternoon

What we got wrong, and what we are saying plainly

We measured late. Our own protocol says re-audit the day each change goes live. The summer and September changes were only scored on October 6, in one block, so we cannot show which of them was worth exactly what. Today's technical package is the first change on this site measured on its own, the same day it shipped.
One point is grading, not progress. Between July and October our checklist tightened its rules. On one technical item the same facts now read as “present but weak” instead of “missing”. We count it in the total because that is how every client is scored today, and we say so here.
Four points went down on purpose. Taking the address, phone numbers and founder names off the site, and putting the top plan behind “contact us”, were our decisions. They cost real points. They stay on the chart.
The second grader wins ties. Every rerun on this site is scored twice — by us, and by an independent model that sees only the evidence and nothing about who wrote the pages. Where the two disagree, the lower score is the one published. Today that took the morning score from 42 to 38, and the afternoon score from 46 to 42 — two static pages were missing a description and a canonical address, which we fixed the same evening.
Our own listing was invisible to our own tools for three days. Google's data feed leaves out businesses that hide their street address unless you ask for them specifically. We fixed the way we ask; the profile had been live all along.

The Cobbler's Shoes, revisited

Boost Systems went from 30 to 42 on its own framework in eleven weeks, and every point is traceable to a specific change on a specific day — including the ones that went the wrong way. The on-page work is largely done: a quotable description, real answers to real questions, prices in text, a technical floor that welcomes every AI crawler. What is left is the part we tell every client is hardest and matters most: existing off your own website. Reviews, listings, and someone else saying your name. That is the next chapter, and it will be measured the same way — one change at a time, scored twice, published either way.