Frequently asked questions

The four blockers:

1. Your site tells AI to go away. There's a file on every website (robots.txt) that says who's allowed in. Thousands of sites have blocked AI crawlers in it without the owner ever knowing. Blocked crawler, no recommendation. Ever.

2. Your text is invisible to machines. If your services, prices, or hours live inside images, PDFs, or only appear after animations load, the AI sees unreadable content: a blank page where your business should be.

3. The internet disagrees about you. Your site says "Dr. Smith Physio," Google says "Smith Physiotherapy Clinic," Yelp has your old phone number. Inconsistent business listings look like an unreliable business to a machine. It recommends someone whose story checks out.

4. Your booking path is a dead end for machines. A broken booking path means the AI can tell someone about you but can't help them become your customer. It prefers businesses where it can finish the job.

The free scan checks your site for all four right now and shows you which ones you have. Thirty seconds, no signup.

There is no submit-to-ChatGPT button. Anyone selling you one is selling air. ChatGPT names whatever it can find, read, verify, and book (booking once agents roll out; the rest is live today) in the seconds after someone asks. Right now, your site is likely failing it here:

1. It can't get in. Your site's access file is blocking AI crawlers, usually without the owner ever knowing. Blocked crawler, no recommendation. Ever.

2. It can't read you. Your services, prices, and hours are trapped in images, PDFs, or sections that load after animations: unreadable content sitting exactly where your business should be.

3. Your story doesn't check out. Your name, address, and phone differ between your site, your Google profile, and your directories. Inconsistent business listings read as doubt to a machine, and doubt gets skipped.

4. Nobody vouches for you. Thin or missing reviews, listings, and mentions on sites that aren't yours. AI trusts the business the rest of the internet confirms exists.

5. It can't finish the job. A broken booking path: forms and buttons a machine can't complete. The AI prefers the business where it can close the deal.

The $99 report finds all five; the $650 fixes clear them, verified on your re-scan.

Old game: you optimized to appear in a list of ten blue links. That was SEO.

New reality: people ask a question and AI speaks one answer, built from the handful of sources it could read and trust. If you're in that answer, you exist. If not, the customer never sees the list you fought to rank on.

AEO (Answer Engine Optimization) is everything that gets you INTO the answer: a site machines can enter and read, facts stated plainly, details that match everywhere, proof from other sources.

It sits on top of SEO. Same foundations, new destination. And it matters because a growing share of searches now end without a single click: the answer box eats them.

That's the entire difference:

SEO: someone searches "physio Montreal," a list appears, you fight for a spot on it.

GEO: someone asks an AI "who's the best physio near me," and the AI names two or three businesses. Generative Engine Optimization is the work of being one of them.

You'll also see AEO, AIO, LLMO. The industry can't agree on a name because the field is under three years old. Which tells you something useful: anyone selling "a decade of GEO experience" just lied to you in the sales pitch.

Skip the vocabulary war. The only question that pays: can AI find you, quote you, and book you (booking once agents roll out)? We fix all three.

Two facts, both measured:

Google is still the front door. It drives the overwhelming majority of search traffic. Being findable there still pays.

The door changed. Roughly 68% of Google searches now end without a click on anything, because the answer at the top of the page absorbs them. Rank #1 under an AI answer and you can still be invisible.

So what died? Keyword-stuffed blogs and bought links, the old retainer playbook. What replaced it: being readable to machines on three fronts at once, Google's crawler, the AI engines writing answers, and the AI agents starting to act for customers.

Three fronts, one pass, flat price. That's the fix.

SEO: rank in Google's list of links.

AEO: get quoted when AI answers a question.

GEO: get named when AI is asked who to hire.

The industry sells these as three separate monthly subscriptions. They are one job done on one website: let machines in, make yourself readable, keep your story consistent, make your booking path work.

One flat fix covers all three. The acronyms are the market's problem, not yours.

What's measured, right now:

AI-generated answers appear across a huge share of Google searches, and independent studies show they cut clicks to the links underneath by half or more.

Younger customers already spread their searching across around five platforms: Google, ChatGPT, TikTok, Reddit, voice assistants.

Nobody serious says Google disappears. The harsher truth is already here: a Google ranking no longer guarantees eyeballs, because the answer box answers first.

Visibility in 2026 means being IN the answers, on every engine, not under them. That's the product.

llms.txt is a simple text file that hands AI systems a clean summary of your site. The idea: make it effortless for machines to understand you.

The honest status, which most shops selling it will never say out loud:

Google has publicly said it doesn't use the file. Most consumer AI engines haven't confirmed reading it either. Some developer tools do use it.

So we install it, because it's free insurance for where things may head. And we put the real work where the payoff is confirmed: access, readable text, consistent details, a bookable path.

A vendor whose pitch centers on llms.txt has a very light checklist. Now you know how to spot them.

When ChatGPT picks businesses to name, it isn't ranking by size, spend, or fame. It's ranking by what it could find, verify, and use in the seconds it had.

A 4-person clinic with a clean, readable site and matching details across the web beats a 40-person competitor whose site is a machine-illegible brochure. It happens constantly.

That's the fairest opening small businesses have been handed in twenty years of internet marketing. It goes to whoever wires for it first, not whoever spends most.

Run the free scan. See if you're currently the readable one or the skipped one.

When someone asks for a recommendation, ChatGPT goes out and reads, right then: your website, your Google profile, review sites, directories, discussion threads.

Then it names the two or three businesses whose story it could read and confirm.

Two consequences that matter to you:

No one can buy placement. There's no ad slot, no submission form, no inside track. Anyone selling one is lying.

Answers vary run to run. It re-reads every time, so the only durable strategy is being permanently readable and verifiable, not gaming one query.

That's a checklist, not a mystery. The free scan grades you against it in 30 seconds.

The gap is almost never quality. It's machine-legibility, and it's invisible from your side:

Their hours and services are written in plain text. Yours are unreadable content: trapped in an image or a PDF.

Their name and phone number match everywhere online. Yours are inconsistent business listings that differ between your site and your directories.

Their booking form is simple enough for a machine to complete. Yours is a broken booking path that defeats it.

Their site lets AI in. Yours has blocked AI crawlers in a file you've never opened.

Every one of those is a technical fix, not a business transformation. One focused pass closes the whole list. That's literally why the fix exists.

Before naming a business, AI checks whether the rest of the internet backs its story: Google Business Profile, review platforms, directories.

Two things follow:

Contradictions cost you. If your site, Google, and Yelp disagree on your name or phone number, a machine reads doubt. Doubt loses recommendations to the competitor whose story checks out.

Reviews are testimony. A business with recent, real reviews on trusted platforms is safer to recommend than a ghost.

The fix includes hunting down every place your business contradicts itself online. The report shows you the list; most owners are shocked by it.

Google ranking and AI citation are cousins, not twins. Google can rank you on years of reputation and links while the AI engines writing answers can't extract a single usable fact from your actual pages.

Result: #1 on Google, absent from every AI answer, and nobody told you.

Why your retainer misses it: the scope was written before answer engines existed. Nobody added the new checks to a contract that predates the problem.

The free scan settles it for your site in 30 seconds. If you're paying monthly for SEO, this is worth 30 seconds.

Independent analyses of millions of AI citations keep finding the same thing: community discussions, video platforms, review sites, and professional networks get cited massively, often ahead of business websites themselves.

Translation: when AI describes your business, it's reading your site AND what the internet says about you.

So the job is both halves. The fix makes your site fully machine-readable, and the report tells you exactly which corroboration you're missing out there: the listings, profiles, and review surfaces that vouch for you.

Anyone selling AI visibility as a website-only job is solving half your problem and billing monthly for it.

Schema markup is labeling code: invisible tags that tell machines "this is our address, this is a price, this is a review score."

The honest evidence, which the schema-as-miracle shops skip: studies tracking thousands of pages found adding markup alone barely moved AI citations, and several engines read only your visible text anyway.

What confirmed wins look like: plain, extractable writing that states facts a machine can lift, with schema as supporting structure underneath.

We implement both, properly. We just refuse to sell the supporting structure as the miracle. This market has enough miracle-sellers.

Agencies: $750 to $2,500 every month. Open-ended retainers, PDF reports about impressions, no end date. The meter IS the business model.

Report-sellers: money for homework. A list of what's wrong. The fixing is your problem.

Sai: $99 for the report, $650 to apply it, sold in that order. Everything found, everything fixed. Optional monitoring at $49 a month, cancel anytime.

No discovery calls. No scoping fees. No invoice surprises. That's the entire price list, and it's public, which tells you something too.

Paste your website into the free scan: 11 automated checks, a grade, and a plain-language list of exactly what's failing: access, readability, consistency, booking path. (The paid fix grades all 24 criteria.)

And know this about the industry: finding what's broken is the easy part, and it should always be free. Scanners like this are everywhere. If anyone is charging you money just to run a diagnostic, or gating a basic scan behind a sales call, they've told you their business model is the report, not the repair.

Diagnosis free. You only ever pay for the fix.

The work starts the moment you complete it, not when we get to it.

Most of the market quotes weeks, because a retainer has no reason to hurry. A flat price does.

The ground under your website moves constantly. AI crawler policies shift. Browser standards update. Platforms roll out changes overnight. The rules that made your site readable this month can quietly change next month, and nobody sends you a memo.

Tracking that is a full-time job. It's ours, not yours.

What Sai actually watches, every month:

Your access files. robots.txt and llms.txt, the two files that decide whether AI systems get in and what they understand once inside. One bad plugin update can silently rewrite them.

Your machine-readable layer. The structure, labels, and technical files under your pages, re-checked against the standards as they exist that month, not as they existed on delivery day.

Your visibility scan. The full rubric, re-run: access, readability, consistency, structure, the works, with your grade tracked over time.

Your booking path. Re-checked monthly: still present, still machine-readable, still wired for the day agents press buttons.

The standards themselves. Crawler policies, browser standards, platform changes, the protocol shifts that never come with a memo. When the rules move, your site moves with them, and that work is included.

The month's result:

Everything passing: one email, one line, ten seconds of your month.

Something broken, because you edited the site, a platform updated, or a standard moved: we re-fix it, included, up to one re-fix a month.

Optional, cancel anytime, one click. The fixes are verified on delivery day; Watch re-checks them while the industry keeps moving.

For scale: monitoring subscriptions in this market run $189 to $489 a month and fix nothing. Ours is $49 and includes the fixing.

The contract is checkable by you: your graded report names the failures, the re-scan shows every one passing, side by side with where you started. Want a third opinion? Run the free scan yourself afterward; it's the same scanner.

The scope is fixed before you pay and the engagement ends when the work is done. No retainer, no renewal, nothing that continues unless you ask for it.

Now the line that separates us from the sharks: nobody on this planet can guarantee your placement inside ChatGPT or Google. Those systems change weekly and answer differently between runs. A guaranteed ranking is a guaranteed lie.

So we don't sell one at all. We sell work you can inspect, on a scan you can run yourself. Companies guaranteeing rankings are guaranteeing something they don't control. Remember that when you read their other claims.

The scam signature, memorize it: guaranteed ChatGPT rankings, secret AI whitelists, miracle files, monthly invoices with no inspectable deliverable. None of those can exist, because no seller controls what AI engines say.

The structure of real work: defined scope, flat price, deliverables you can inspect, results you can verify yourself.

We built the offer so you never take our word for anything: free scan before any money, itemized fixes, and a re-scan you can run yourself.

Scams don't survive that structure. That's why they don't use it.

AI engines change weekly. They weigh sources differently per question. They answer differently between two identical runs. Guaranteed placement inside them is a guaranteed lie, and every reputable operator in this industry says the same thing in unison.

What CAN be checked is the state of your website: machine-readable, consistent, corroborated, bookable (booking wired now, live when agents roll out). We verify exactly that on your re-scan, and you can run the same scan yourself.

Buy outcomes from the only category that can deliver them: the work itself. Anyone selling the other kind just told you who they are.

This work didn't exist when your site was built. It moved three times this year alone. Your developer builds and maintains websites; tracking how five different AI engines read, verify, and act on them is a different full-time job.

The practical math: at $99 for the report and $650 to apply it, we're under most developers' day rate, we arrive with the current checklist, and we leave behind a verified before-and-after your developer can't produce.

No knock on your web guy. It's a new job description, and it's ours.

The agency model is antiquated, and you fund it: office overhead, account managers, sales teams, and a retainer structure where the meter running IS the product. $750 to $2,500 a month buys you a seat in that machine, and the machine needs the work to never finish.

We built the opposite machine:

Our own scanning technology. Sai, our own agent, runs the diagnostics and the monthly monitoring. No licensed tools marked up, no junior staff clicking through checklists.

Our own delivery stack. The rubric, the report, the wiring, the verification pipeline: built once, in-house, run per client at almost no overhead. That efficiency goes into the price instead of an office lease.

A model with an end date. Fixed scope, executed by a person, verified on the re-scan you run yourself. We get paid to finish, not to remain.

This isn't the discount version of a retainer. It's what the same technical work costs when nobody's business model depends on stretching it across a year. You're not buying less. You're buying a company that doesn't need you trapped.

The normal way: every platform (WordPress, Shopify, Wix, Squarespace) has a built-in feature for adding an extra user. You add us as a temporary user with limited permissions, takes about two minutes, we send you the exact clicks. We do the work under that account. When we deliver, you remove the account, one click, and we can never get in again. Your own username and password are never shared, never seen, never asked for.

The no-access way: if you'd rather nobody outside touches the site, we do the work on paper instead: every fix delivered as ready-to-install files with exact instructions, and you or your web person applies them.

Either way, what shows up in your inbox is the same: the fixes, the report, the proof.

One warning worth keeping: any vendor who asks you to email them your actual password just told you everything about how they run their shop.

What the fixes physically are: machine-readable structure added underneath your pages, corrected technical files, access permissions for AI crawlers, wiring on booking paths that already exist.

What they are not: design changes, content rewrites, hosting moves, or anything a returning customer would notice.

Additive, standards-based work doesn't trip Google penalties; it's the structure Google's own documentation asks for.

And it's all verifiable, which is the point: you get the before-and-after in your re-scan, so trust is never the required ingredient.

Contracts exist in this industry to protect revenue from results. Lock the client for twelve months, and the work can afford to be mediocre by month three.

We inverted it: the fix ends, Watch survives only as long as it's obviously worth $49 to you, and cancelling is a click, not a conversation.

We'd rather be re-hired than legally retained. It keeps us sharp and it keeps you free. That's the whole policy.

This industry is drowning in invented vocabulary: GEO, AEO, AIO, LLMO, miracle files, secret AI whitelists. Most of it exists to make ordinary work sound proprietary and to make you feel too far behind to ask questions. We don't play that game, because a customer who understands the work is exactly the customer we want.

So, plainly:

The part that overlaps with SEO, and we say so: clean technical structure, readable content, consistent business details. If your site already has perfect technical SEO, part of our checklist is already done. Most sites don't.

The part that's genuinely new, number one: AI engines now write answers and cite the sources they trusted. Getting INTO those answers has its own checks: crawler access rules, extractable text, corroboration. Your old SEO scope doesn't contain them, because it was written before answer engines existed.

The part that's genuinely new, number two: AI agents are starting to act on websites, filling forms, booking (rolling out now), requesting quotes. Wiring your booking path so a machine can complete it is work no traditional scope has ever included.

What physically changes on your site: files edited, structure added underneath your pages, access rules corrected, your booking path wired. Nothing your customers see. Everything a machine reads.

That's the whole answer, no vocabulary required. Any vendor who can't explain their service this plainly is hoping you won't ask.

Two Shopify facts:

The default machine-readable data covers only basics. Product name, price, and not much else. The details that win recommendations sit unlabeled.

Shopify switched on agent-facing infrastructure across its stores this year. The platform is preparing for AI shoppers at full speed, which makes the gap between "wired store" and "default store" wider every month there.

We extend what Shopify starts and wire what it doesn't. Same $99 and $650, same proof.

Both platforms accept the fixes through their code injection features. The work applies.

The honest platform note you won't find in a sales proposal: site builders limit how deeply their built-in machine-labels can be customized, and content hidden inside tabs and accordions can be invisible to AI crawlers entirely.

The report flags plainly if your platform is capping anything. You'll know your ceiling before a dollar of work starts.

Full control over structure, labels, technical files, and wiring. The entire checklist applies without platform limits.

If your site is WordPress, the only question left is what the scan finds. Thirty seconds to find out.

What it means, physically: right now, an AI assistant visiting your website is a tourist without a map. It looks at your page, guesses which button books an appointment, guesses what your form wants, and often guesses wrong and gives up. WebMCP is a short piece of code on your site that replaces the guessing with instructions: this is the booking action, these are the fields, this is how you complete it.

So what? Two things:

When someone's AI is sent to book a service (rolling out now), it books where it can succeed. A site with WebMCP wiring is a door that opens. A site without it is a door the AI rattles and leaves. The customer never sees any of this happen; they just get booked somewhere, and it's either you or it isn't.

The wiring is nearly free right now, and it won't stay optional. It rides along inside the fixes at no extra cost, a few hours of work while we're already in there. The businesses that added mobile-friendly early paid little and led; the ones who waited paid more, under pressure, behind competitors. Same curve, earlier point on it.

Status: testing is ongoing in Chrome, with major platforms already building for it. It's on track to become the standard for how AI interacts with websites.

So what does that mean in money:

Door one, Google: paying today. Roughly 68% of searches end in the answer box, not a click. If your site isn't readable to the machine writing that answer, you lost those customers this morning.

Door two, AI recommendations: paying today. People ask ChatGPT and Perplexity who to hire, every hour, right now. Those engines are naming businesses in your city, in your category, this week. Either yours is namable or a competitor is getting introduced to your customers by a machine they trust.

Door three, agent booking: getting wired while we're already inside. Google and Microsoft built the standard. Shopify and Cloudflare switched on the infrastructure. The wiring costs you nothing extra inside the fix.

That's the whole answer. You pay $99 for the report and $650 to apply it for the two doors bleeding today, and door three comes bolted on for the day it opens.

Mobile-friendly, 2012. Early wirers paid little and led for a decade. Late wirers paid rush prices to catch up to a line that had moved without them. Same curve. This is the cheap end of it.

llms.txt is a plain text file that sits at one fixed address on your website (yoursite.com/llms.txt). Inside it: a clean, organized summary of your business written for machines. What you do, where you are, which pages matter, where your services and prices live.

Think of the pair like this: robots.txt is the bouncer that decides which machines get in. llms.txt is the menu you hand them once they're inside, so they understand you in one read instead of piecing you together from forty pages.

Where it stands: it's a proposed standard. Some AI tools read it today. Google has said it doesn't. More engines picking it up is the direction things point.

Your services, prices and hours can sit trapped in images, PDFs, or sections that only load after animations. A person reads them without noticing anything is wrong. The crawler writing the AI answer sees unreadable content sitting exactly where your business should be.

The fix is not a redesign and it does not change what your visitors see. Those facts get put into text a machine can lift, as machine-readable structure added underneath your pages, so the page you already show a customer is a page an engine can read.

Readability is one of the things the free scan reports, alongside access, consistency and your booking path. The paid audit grades it among all 24 criteria.

A site can register actions in JavaScript that only appear once the page runs, and this scan will not see them.

A finding here is a sign to check, not proof that something is missing.

What FAQ schema actually is: invisible labeling code wrapped around your questions and answers, telling machines "this is a question, this is its answer." For years, Google rewarded it with expanded listings in search results, so the whole industry bolted it onto everything.

Then two things happened:

Google retired the reward. FAQ rich results are gone as of this year. The expanded listings that justified the gold rush no longer exist.

Independent testing exposed the markup. Studies tracking thousands of pages found adding the labeling code alone barely moved AI citations, and several engines read only the visible text on your page, ignoring the invisible labels entirely.

We still add the markup where it belongs, because it's minutes of work and correct structure never hurts. But any vendor still selling FAQ schema as the product is selling you 2019.

Clock one, the work. Every fix applied and verified passing on your re-scan. This clock we control completely.

Clock two, the engines: 2 to 4 weeks for first movement, about 90 days for a real footprint. AI engines revisit and re-verify on their own schedule. Nobody controls that clock, and honest operators say so.

We set your baseline on day one, so movement is counted, not felt. See question 44 for the four numbers.

1. Open a fresh, logged-out ChatGPT session (your own account skews results).

2. Ask: "best [your service] in [your city]." Note who gets named.

3. Ask about your business by name. See what it knows and whether it's accurate.

4. Ask a buying question in your niche: "who should I use for X near me."

5. Repeat each a few times. Answers genuinely vary run to run; a pattern is data, one screenshot is nothing.

6. Run the same on Perplexity, which shows its sources, so you can see exactly which sites earned your competitors their spot.

That's your baseline. It's also exactly what Sai Watch does on schedule so you never have to remember to.

There is no rank inside ChatGPT to report, so honest measurement is:

1. Citation rate. How often AI answers cite your site on your key questions.

2. Mention rate. How often your business gets named, cited or not.

3. Citation share. Your share of those citations versus your competitors, same questions.

4. AI referral traffic. Visitors arriving at your site from AI engines, counted in your analytics.

Baseline on day one, re-measured on schedule, delivered in one plain-language email. That's the whole reporting stack, and it's four real numbers instead of one invented one.

Here's why those questions are the ones everyone's asking: when someone asks an AI who to hire, it names two or three businesses. Those two or three get the call. Everyone else lost a sale they never knew existed. No missed-call log, no bounced visitor in your analytics. The customer was decided before you knew they were shopping.

Sai is how you become one of the names it says. Found by ChatGPT. Found by Gemini. Found by Perplexity. Verified on Google. Readable, consistent, bookable (booking wired now, live when agents roll out), verified on your re-scan, measured from a day-one baseline.

Get named, and the sales follow the same math they always have. That's the job, and Sai does it.

AI engines read your French and English presence as ONE business. French site, English Google listing, mismatched details between them: a machine reads doubt, and doubt loses the recommendation, in both languages.

There is no submit-to-ChatGPT button, en français ou en anglais. There is being readable, consistent, and corroborated across both languages simultaneously.

That's the fix, delivered bilingually: service, report, and proof in French or English. Few providers do this seriously in Quebec. We're based in Montreal. C'est notre marché.