THE 45-SECOND CASE · WHY THIS SUMMER CHANGED THE STAKES
CRAWLER RULE CHANGE · SEPTEMBER 15, 2026
Cloudflare announced new AI traffic controls on July 1, 2026. From September 15, multi-purpose crawlers are evaluated on all their behaviors — which means a legacy “block AI bots” rule can also block Googlebot, Applebot, and Bingbot.1 Thousands of businesses flipped that switch in 2024 and never looked at it again.
Meanwhile, Google is rolling out AI that calls and books local businesses on customers' behalf — starting with home repair, beauty, and pet care — across the U.S. this summer.2
Translation: machines are becoming your busiest visitors. Most websites weren't built for them. We audit yours in three scores and tell you exactly what to fix first.
PHASE-BILLED ENGAGEMENTS · NO LONG-TERM CONTRACT REQUIRED TO GET YOUR SCORES · SERVING 8 LOCAL-SERVICE VERTICALS PLUS B2B MANUFACTURING · NATIONWIDE
I WHY NOW
You don't need to take our word for any of this. Both changes are published, dated, and already in motion.
EFFECTIVE SEPT 15, 2026
Cloudflare changes how crawlers are judged. Multi-purpose crawlers get evaluated on all behaviors — so a blanket “block AI bots” setting from years past can now catch Googlebot, Applebot, and Bingbot in the same net. If that's you, you could vanish from ordinary search, not just AI answers.1
SOURCE: CLOUDFLARE AI TRAFFIC CONTROLS, ANNOUNCED JULY 1, 2026
ANNOUNCED MAY 19, 2026 · ROLLING OUT NOW
Google's AI is calling businesses for customers. At Google I/O, Google announced agentic booking and AI calling on users' behalf in home repair, beauty, and pet care — U.S. rollout summer 2026. When that call or booking attempt hits your business, it either completes… or it moves to a competitor.2
SOURCE: GOOGLE I/O KEYNOTE, MAY 19, 2026
Google's own developer guidance now tells site owners to build agent-friendly websites3 — and Chrome's Lighthouse tool added an Agentic Browsing category in June 2026 (currently informational and unbenchmarked) to measure exactly this.4 The tooling to grade your site already exists. The only question is whether you see your grade before your competitors see theirs.
II THE 90-SECOND FAILURE
A homeowner's AC dies on a Friday night. She asks her AI assistant to find someone good nearby. The model answers — and it names your company. Your reviews, your service area, your years in business. You won the citation. Your marketing worked.
9:42 PM — THE HANDOFFShe says: “Book the first slot tomorrow.” Now a different actor takes over — not the model that answered, but an agent acting in real time on her behalf. It loads your website looking for something it can act on.
9:42 PM — THE WALLIt finds a phone number embedded in an image. A contact form with unlabeled fields. Service hours in a PDF. Nothing it can parse, nothing it can book, nothing it can confirm. It can see a brochure. It cannot see a business.
9:43 PM — THE LOSSThe agent moves to the next name in the answer — the competitor with labeled forms, structured data, and a bookable path. The job books there. You never see it happen. No missed call, no lost form fill, no trace in your analytics that reads as anything but a bounce.
The gap between “the AI named you” and “the agent booked you” is about ninety seconds — and it's where local jobs are quietly changing hands.
III THE FIX
Most agencies sell you one number and call it “visibility.” We score three separate questions, because three separate actors decide whether you get paid.
PILLAR 1 · FIND
Can an AI system find and cite you? When the model composes its answer, does your business make the list?
DECIDED BY: THE MODEL, AT ANSWER TIME
PILLAR 2 · CONVERT
Can a human convert once they land? Clear offer, fast page, obvious next step, no friction.
DECIDED BY: A PERSON, ON YOUR PAGE
PILLAR 3 · TRANSACT
Can an agent act — book, quote, buy — on a person's behalf? Parseable data, labeled controls, a completable path.
DECIDED BY: AN AUTONOMOUS AGENT, IN REAL TIME
GEO gets you named in the answer. AR is what happens ninety seconds later when the customer says “ok, book it.” A business can score well on one and fail the other — which is exactly why we never blend them into one vanity number.
IV DELIVERABLE 1
Our flagship audit. Three independent scores — 8 scored categories per pillar — plus a prioritized fix list in plain English. Here's what a typical result looks like:
| SCORE | BAND | WHAT IT MEANS |
|---|---|---|
| 85–100 | AGENT-TRANSACTABLE | Agents can complete a booking, quote, or purchase end-to-end. |
| 70–84 | AGENT-LEGIBLE | Agents understand you but hit friction before the finish line. |
| 50–69 | AGENT-VISIBLE | “Agents see a brochure, not a business.” |
| 30–49 | AGENT-BLOCKED | Access or structure failures stop agents cold. (Sample above: 34.) |
| 0–29 | AGENT-INVISIBLE | To an agent, your business effectively does not exist. |
V DELIVERABLE 2
The AI Visibility & GEO engagement. Highest-risk items first: the very first task is making sure you're not blocking legitimate crawlers and agents before the September 15 change.
1
Technical baseline plus the AR-1 agent-access unblock: crawler rules, bot settings, and access controls corrected so you're not fighting the Sept 15 crawler-evaluation change.
AR-1 DONE IN WEEK ONE
2
The pages, structured data, and answer-ready content that get you found and cited — built for models and humans at once.
3
CRO and agent UX handled in one dev pass: the same labeled forms, clear paths, and parseable actions serve human visitors and autonomous agents alike.
4
Quarterly re-scores and call testing, so you see your three scores move — and catch regressions before they cost bookings.
VI WHO THIS IS FOR
Google's first agentic-calling wave covers home repair, beauty, and pet care.2 If you're in one of those three, this is not a someday problem — the calls are starting this summer.
AGENTIC WAVE 1
HVAC, plumbing, electrical, roofing. AI is already being positioned to book the emergency call — the exact 9:41 PM scenario above.
AGENTIC WAVE 1
“Beauty” is named in Google's first calling wave. High-frequency rebooking makes agent-completable scheduling a compounding advantage.
AGENTIC WAVE 1
Pet care is in wave one. Anxious owners delegate to assistants; the clinic an agent can actually book wins the panic-hour appointment.
Insurance questions plus scheduling — a natural fit for structured, agent-readable answers.
Class schedules and intro offers that models can cite and agents can act on.
Quote requests and drop-off booking, structured so an assistant can complete them.
AI visibility with advertising-compliance guardrails — see our state ad-compliance audit below.
If customers find you by asking questions and book you through a website or a phone line, the three-score model applies.
VII B2B MANUFACTURING
Manufacturing is structurally different, so we audit it differently. A sourcing agent working a bill of materials doesn't browse your site — it interrogates it. When it can't extract an answer, you're not shortlisted, and no human ever knew you were considered.
What an agent-ready manufacturer looks like:
The scorecard's AR pillar swaps consumer-booking checks for sourcing checks: can an agent qualify you, spec you, and start an RFQ without a human on your side lifting a finger?
VIII ALSO ON THE SHELF
FOR REGULATED ADVERTISERS
Legal-advertising compliance review for regulated verticals built to your state’s rulebook. Know your marketing meets state advertising rules before a regulator, a bar association, or a competitor points out that it doesn't.
PRODUCTIZED RESEARCH
Fixed-scope, fixed-fee research products. Order one, get a report:
IX BEFORE YOU EMAIL
Every engagement is quoted up front and billed phase by phase — you can stop at any phase, and each phase ends in a verifiable deliverable. No per-lead, per-booking, or percentage-of-revenue arrangements. In regulated verticals like legal, outcome-tied fees can create compliance exposure for you, not just us, so we don't use them. You'll know the cost of each phase before you say yes.
No — and you should be suspicious of anyone who does. AI systems change weekly and no vendor controls their output. What we commit to is defined scope, dated deliverables, and quarterly re-scores so you can see the trend line for yourself and hold the work accountable.
SEO optimizes for a results page a human scrolls. GEO optimizes for an answer a model composes — and AR optimizes for an action an agent completes. Your SEO agency has probably never tested whether an autonomous agent can book you. That third pillar is the difference, and it's the one with a September deadline attached.
We track them, we don't oversell them. As of June 2026, about 8.7% of the Tranco top 1,000 sites had adopted llms.txt (Rankability), and Google states its Search doesn't use these files.5 It's a low-cost line item in Phase 2 — never the headline fix. The headline fixes are access, structure, and a completable path.
Google's web.dev guidance on building agent-friendly websites (updated April 2026)3 and Chrome Lighthouse's Agentic Browsing category (announced June 22, 2026, Chrome 150+ — currently informational and unbenchmarked)4 define the emerging baseline. Our AR pillar scores 8 categories against that guidance plus live testing: we attempt the paths an agent would attempt and document where they break.
You get a reply with a short intake (5 questions), a fixed quote, and a start date. The scorecard lands as a written report with your three scores, the 8-category breakdown per pillar, and a prioritized fix list. If you stop there, you still own a complete diagnosis. If you continue, Phase 1 starts with the AR-1 agent-access unblock in week one.
X THE ONLY NEXT STEP
One email starts it. You'll get your GEO, CRO, and AR scores, the band you're in, and the ordered fix list — before the agents come calling.
Or just email info@cortexnavigation.com and say “audit.” We'll take it from there.
OPEN WATER · SERVING ALL 50 STATES
All scores shown on this page are labeled SAMPLE and are illustrative, not client data. WebMCP is referenced as an emerging standard currently in a Chrome 149 origin trial; it has not shipped.