Your Google ranking is rotting. ChatGPT doesn't cite you.

The breakage (what's actually happening)

Buying decisions used to start at Google. They now start at the answer engines — ChatGPT, Perplexity, Gemini, Claude, Google's AI Overviews. When a buyer asks "best [your category] in Houston" or "what's the right peptide stack for X," an AI answers, and it names 2-4 sources. If you're not one of them, you don't lose a click. You lose the buyer. They never reach your site.

Here's the part most SEO is built to ignore: ranking on Google and being cited by an AI are not the same thing, and the gap is widening.

• The overlap between top-10 Google rankings and AI-Overview citations collapses from ~76% down to ~38% as query volume scales. The rank you spent years buying is decaying against the engine that actually answers your buyer.
• For AI citation, brand mentions outperform backlinks roughly 3:1. Unlinked brand mentions correlate with citation at 0.66–0.71; backlink count sits around 0.22. Most of the SEO industry is still optimizing the weaker signal.

You are, almost certainly, pouring budget into the wrong engine.

The proof (why this is different from the 400th AEO newsletter)

Two things separate this from the wave of "AI SEO" hot takes:

1. Measured research, not vibes. The numbers above aren't opinion — they're from a citation-correlation study across live AI-Overview output (YouTube presence 0.737, unlinked brand mentions 0.66–0.71, branded search 0.35–0.47, domain rating 0.27–0.33, backlink count ~0.22). The implications are counter-intuitive and load-bearing: brand mention velocity beats link-building for AI visibility. Most agencies haven't noticed.

2. The only causal measurement discipline in the category. Everyone else reports observational correlations ("sites with X tend to get cited more"). Correlation isn't causation — and in a nascent category, buying a $30K retainer on correlational evidence is a gamble. The method that actually proves a lift is a prospective intervention/control forward-test: freeze a query set, split it 15 intervention / 15 control, optimize only the intervention set, wait six weeks, re-measure citation visibility, compare. That's the design. It's the difference between "we think this works" and "we can show you it worked, on your queries, against a control."

This is the methodology baked into how the system is run — and into the pilot below.

The system (Constellation)

A multi-model agent pipeline, not a consultant with a deck.

• Speed: sub-5-second quick checks from the edge; sub-30-second deep analysis; ~1,000 sites/hour at scale.
• The AI Discoverability Score — four weighted pillars: Technical Compliance, Content Structure, Entity Authority, Voice AI Readiness. One number, decomposable into the exact gaps causing it.
• Validity-gated, deterministic reports — every score carries provenance and a degraded-flag if the evidence is thin. No LLM-confident-wrong summaries. A report that says PASS actually asserts on content, not on a status flag.
• The scorer is calibrated to citation, not to rank. llms.txt capped at ≤2% weight. Content-length floors raised to 1,200–2,500 words. The GEO composite split into citation-selection vs. citation-absorption dimensions (grounded in arXiv 2604.25707) — because being picked and being quoted are different problems.

The offer — audit → pilot → retainer

No big-bang engagement. Risk reverses at every step.

1. The Audit — $15–18K (yours to keep)

Full AEO/GEO scan of your domain. You get the AI Discoverability Score, the decomposed gap map across the four pillars, and a prescriptive fix list prioritized by citation impact. Delivered as a report you can hand to any team — or to us. This is the wedge: it proves we understand your specific visibility problem before anyone commits to anything.

2. The Pilot — $4,500/mo, scoped to one property

We close the top gaps on a single property and run the causal measurement on it: the frozen query set, the intervention/control split, the six-week re-measure. The deliverable is the measured citation lift on your domain against a control — not a slide that says "trust us." If the intervention set doesn't move against control, you don't enter the retainer. This is the "$100 bill for $20" moment: the pilot pays for itself the moment the lift shows up in the queries your buyers actually ask.

3. The Retainer — $27–35K Phase 2 / ongoing

Operate the system: refresh against model and research drift (models upgrade quarterly; the scorer gets conflict-resolved against new deep-research), expand across your property portfolio, and keep the citation measurement running as the standing scoreboard. This is the part most consultants skip — an AEO position drifts the moment nobody owns it.

Real anchors, not theory: a live $4,500/mo AEO pilot on sermorelin.com (Halo Brands' 100+ peptide-domain portfolio); a scoped $15–18K → $27–35K phased engagement structure; an $8.5K build + $1,500/mo retainer shape on prior content-automation work.

Who this is for

Operators who hold a domain portfolio or a category where the buyer's first question now goes to an AI: specialty law firms, peptide/pharma domain investors, hospitality groups, indie SaaS founders losing organic to AI Overviews. Principals who recognize the moat is technical and hire on proof, not pitch.

Not for: anyone who wants a charismatic front person, a deck without a codebase behind it, or "AI automation" resold from someone else's templates.

The call (not "buy now")

Twenty minutes. We run a live quick-check on your domain on the call — you see your AI Discoverability Score and the top three gaps in real time, before any conversation about engagement. Then you decide.

Reply to this message or email aria@sorcerai.co with "AEO diagnostic" to book a 20-min live quick-check on your domain.