Product Updates

The biggest week we've shipped. The front door got a second key, a new account now starts with its score instead of a button, and two public datasets went live: the AI Citation Index and Brokerage Rankings.

Create your account with an email address

Every place the product offers "Sign in" now offers the same three doors: an email address and a password, Google, or Apple. Pick the one you want on day one and it works everywhere afterward — the report, the dashboard, the intake, the checkout. Forgot the password? A reset link goes to your inbox and brings you straight back. And the free path from your report lands directly on your dashboard.

Your dashboard opens with your score already on it

When you create an account from your free report, the dashboard's first load shows that score — not a "run your first score" button pointing back at the page you just came from. And the moment the account exists, GeoGenius asks ChatGPT, Grok, Google AI, and Claude about you. The Visibility tab reads "Asking ChatGPT, Grok, Claude, and Google AI about you…" while it runs and shows the answers when it lands. Nobody has to find a button, and it runs for every account, free tier included.

"Finish your profile" now describes exactly what's left to do.

The AI Citation Index is public

/agents is the record of which real estate agents ChatGPT, Claude, Gemini, and Grok actually name when asked for an agent in a US market. We ask the engines the questions a buyer or seller would ask, store every answer, and list an agent once two responses — or two different engines — have named them. Browse by state to a metro, then to the agent: the engines that named them, how many responses, first and last citation, and their website. Lists are ordered by evidence; payment plays no part, and contact details are never published.

The engines are re-asked on a rolling schedule, so the record stays current, and every metro is available as JSON for anyone who wants to check our work. If you're on one of these lists, claiming your entry links it to your profile.

Brokerage Rankings

/brokerages rolls the same record up to the brand. In every ranked metro, the agents the engines name are grouped by the brokerage they work for — Keller Williams Realty Bellevue, KW Eastside, and a team at Keller Williams SouthPark all count as Keller Williams — and ranked by named agents, then distinct responses, then distinct engines. National, by state, by metro, and one page per brand.

It measures presence in AI recommendations, not sales volume, headcount, or quality: a large brokerage with few AI-named agents ranks below a small one whose agents the engines recommend. A metro is ranked once at least five named agents span three or more brokerages.

The report now looks at what crawlers meet in front of your site

A new section on the score report, "What AI crawlers meet in front of your site", shows the layer most agents never see: whether your host or CDN challenges bots, how your site answered each crawler's user agent, whether pages arrive server-rendered or as an empty shell that only a browser can fill in, and what your robots file asks of crawlers. It's split by who can fix it — your host or platform, or your site template and content — because that's the first question a fix raises.

These findings are diagnostics rather than score points. They exist because on-page work can't be read by an engine that gets turned away at the door.

A week on the dashboard. The engine check stopped being a one-off and became a record, the tabs were reorganized around how the engines actually judge an agent, and a list of dashboard details got the attention they needed.

The engines are asked about your niche, and the answers are kept

The four-engine check now asks three questions: a fixed market benchmark — the same question as our 36-market study, so your result is comparable to the field — plus two composed from your profile: the niche and ideal client the intake now captures, and what your pins and transactions say about where you work. A composed question can never contain your name, team, brokerage, or website; that would be asking the engine to cheat.

Your questions are frozen once written, so week-to-week results mean something, and they're only re-tuned when your niche or market changes or enough new pins and transactions accumulate to sharpen them. Every run is kept. The dashboard charts how many of the four engines named you, per question, over time, with markers where the questions were re-tuned and the wording of each one in the legend. On paid tiers the check re-runs itself weekly.

The dashboard shows a running check as it happens

When a check is running — including the one that starts the moment your account is created — the dashboard shows it in progress, fills in the engine cards and the chart as the answers land, and won't start a second run over the first.

One dashboard, organized the way engines grade you

The tabs now follow the framework the engines are graded on: Overview · Visibility · Knowledge · Trust · Chat · Leads. Visibility holds your report and every on- and off-site fix, with the four engine answers alongside it instead of on a tab of their own. Knowledge is the first-hand work: the map, your pages, pins, and transactions. Trust is new — testimonials, profile completeness, connected sources — because other people's words about you aren't knowledge you hold.

Your own dashboard shows the owner's view of your report — your results, with no signup steps or offer blocks. And "Get the fix" now lands on the fix: each Fix Pack fix sits under the finding it answers, and a plain checklist of every fix — worst first, one line each, code on demand — lives beside the report.

Also this week

• The Fix Pack download works on every paid account, with a clear message and a retry if generation ever needs a second attempt.
• Large transaction uploads finish: the browser waits as long as the server does, and geocoding runs six rows at a time.
• Profile edits from the dashboard save, with a test that guards every place the call is made.
• Re-scanning your own site while signed in keeps the report attached to your account, so the Fix Pack always finds it.
• New deploys land on your next page load.
• One "Re-run scan" button, on the Visibility tab; suggestions open details and never start a scan.
• The map centers on your area when you have no pins yet, and the account page shows and edits your full profile, website, and email.

From the blog

• AI Visibility Score for Real Estate Agents: What It Measures and How to Raise Yours — the 11 dimensions, in the order to work them.

The densest week so far. The capture flow got the plumbing it needed to be trusted in the field, the report learned to see everything that lives off your website, and a week of testing on real agents' sites made the scan sharper.

Your neighborhood guide now grows with every pin

Every pin now refreshes the guide for the neighborhood it was dropped in, so a guide keeps growing as long as you keep working the area. Pins are filed by geography — a pin dropped standing in Elmhurst about the school two streets over belongs to the Elmhurst guide whether or not you said the name out loud.

A burst of pins on one tour turns into one rewrite, not a dozen. Guides are written in your voice — the guide prompt shares the pin prompt's ban on brochure adjectives — and every answer in a guide's FAQ has to be supported by one of your pins.

A recording survives a dead signal

A pin spoken in a new build with one bar of signal is saved to your phone before anything else is attempted, and it's cleared when the pin is saved — not when the transcript comes back — so closing the app at the review step loses nothing. A pending strip shows recordings still on their way; it takes no space when there's nothing waiting.

The review screen now tells you what a pin builds: "This updates your Elmhurst guide" before you save, and a link to the guide after. It will never link you to a page that isn't live yet, and it never claims a rewrite that hasn't happened. Tapping a saved pin reads "Cougar Mountain · 5 pins · has a guide" or "Montreux · 1 pin · 2 more for a guide." Pins dropped before we recorded neighborhoods say so instead of guessing. And a tapped landmark now shows its name and the area it files under — "Lewis Creek Visitor Center · Cougar Mountain" — instead of a postal address.

The dashboard on your phone is built for capture

On a phone, the dashboard stops being the desktop shrunk. Top to bottom: your score at a glance, a capture card with one big "Drop a knowledge pin" button, anything still pending, and a neighborhood scoreboard — every area with its pin count, the ones a single pin short of a guide on top ("1 more pin creates your Montreux guide"). Pin ideas appear where an area is missing a category: "Cougar Mountain has no school pins yet — what do parents ask you about the schools?" Desktop is unchanged; a narrowed desktop browser still gets desktop.

The report sees off-site now

The Fix Pack and the report stopped at your website. Now they cover what lives off it. New coded findings for Google Business, brand authority, and review strength, each with a remediation that never promises review automation or invents a number. A content studio drafts what those fixes need — Google Business description options, review-reply templates, request scripts, a four-week post calendar — with every draft checked against your real data before you see it.

Two new sections on the free report. "Where the engines get their answers" ranks the sources ChatGPT, Grok, Google AI, and Claude cite for agents in your market and shows whether you're present on each, with any name, phone, or city conflicts between your site, your Google profile, and the portals spelled out with the exact values. "Who the engines recommend in your market" names the agents the engines currently recommend, which engines said so, and a row for you. We also check whether your name appears in Wikidata, whether your structured data has a stable identifier, and whether your name is spelled the same everywhere. Where a source blocks our check, the report says "couldn't check" — never "absent."

Sharper scans on real sites

A week of scanning real agents' sites, and five improvements from it:

• Market-intelligence stages run in parallel with a hard time budget, and a scan that can't finish says so and offers a retry.
• When a host challenges our probe, the report says so, awards no points either way, and tells you exactly what to allow at your CDN.
• Structured data is parsed the forgiving way the engines parse it, so a stray line break inside a value never hides the whole block.
• Phone checks compare every number on a page, so a site with two phone numbers in its header is never flagged against itself.
• Anonymous scans earn NAP Consistency points from public sources: name coherence, the phone on your Google profile, and your city.

A week split between the sidewalk and the library. Knowledge pins can now be captured by voice from your phone, and the guides were rebuilt from sources.

Drop a knowledge pin by voice, from where you're standing

Open GeoGenius from your home screen, and the app opens on a map centered on where you are, with your existing pins around you. Drop the pin, talk for twenty seconds about the block you're standing on, and you get a saved knowledge pin back in your own words. The cleanup step removes filler and fixes grammar, and it is forbidden from adding facts, marketing adjectives, third-person voice, invented numbers, or calls to action. Your raw transcript is always shown alongside, with a one-tap "Use my exact words."

If a new pin sounds like one you've already dropped nearby, the app says so and offers to keep both, keep the new one, or keep the earlier one — quoting what would be lost, and never deleting without asking. On phones, every "Add a pin" button now opens this flow. Desktop is unchanged, and the app icon is now the GeoGenius teardrop rather than a letter in a box.

The guides were rebuilt from sources

All six guides are new or rewritten: Google Business Profile, Foursquare, reviews and testimonials, knowledge pins, transaction history, and how GeoGenius builds your neighborhood guides and sold pages. Every number in them traces to our 36-market study, and each guide ends with the questions agents actually ask, answered.

Pages built for crawlers now reach them

The guides are served as real HTML, so an AI crawler that doesn't run JavaScript reads the whole guide. Prerendered pages answer at their clean URL with no redirect hop, and the sitemap lists every live blog post with its real publication date.

From the blog

• Where AI Search Is Heading in Real Estate: What the Evidence Shows — the trajectory, argued from public data rather than predictions.

The biggest week since the score itself. The report has always told you what's wrong and stopped there. Now there's a file that fixes it, and the scan underneath it got sharper.

The Fix Pack: the file that fixes what the report found

The AI Visibility Fix Pack is one downloadable work order: every on-site issue the scan found, the exact change for each, the evidence from your own scan, the code to paste, where it goes on your website platform, and how to verify it worked. A JSON version ships alongside for anyone with a developer or an AI coding tool. Hand it to whoever runs your site, or work through it yourself.

The line between free and paid: the free report gives you your score, what all four engines say, and the full ranked list of what's wrong. Visibility — $245 for six months, or $49 a month — gives you the file that fixes it, with a money-back guarantee on the on-site points, because those are the ones our own engine measures end to end. We make no timing promises: six months is how long we work, not when results arrive. There's an annotated specimen of the file on /report.

The report is organized by where the work lives

Every score breakdown now splits on one axis. On your site, 50 points: Structured Data, Crawler Access, Content Citability, NAP Consistency, Website Quality — exactly what the Fix Pack covers. Off your site, 50 points: Brand Authority, Google Business, Knowledge Depth, Transaction Authority, AI Engine Visibility, Review Strength. "Needs an account" is now a label on a line, not a bucket of its own.

The scan reads your full page

The scan now reads the raw page — head section, scripts, and footer included — so structured data, meta tags, and footer links count wherever they live on your site. Link checks match real links rather than the word "Zillow" in a paragraph; robots.txt, your sitemap, and llms.txt are fetched directly; page load time is your site's own measured timing; and anything we can't verify is reported as "could not be verified" — never as absent. Scores move on fresh scans as a result: up where signals are now detected, and down where Crawler Access points now require evidence.

The Content Citability suggestion now points at the Fix Pack, where that work lives (pins feed Knowledge Depth).

Finish your profile on one screen

After the report, the seven-step chat is gone. You sign in and see every field we pulled from your site, pre-filled and editable, on one screen. Name, city and state are the only requirements; submit, and you land on the dashboard while your bio and the engine checks run in the background. If you'd rather talk it through, the chat is one link away.

A short week, spent on giving the report a page of its own and tying every page that describes it to the engine that computes it.

/report shows you the real thing

What's in the Report is new: a page that answers "why run a GeoGenius report, and what do I get out of it?" Its centerpiece is an annotated live sample — the actual score ring, source scan, category breakdown, and metro curve rendered from sample data, with callouts explaining each part. It is the real artifact, not a screenshot, so it can't drift from the product. Below it: what the score measures, what you do with it, where you stand against agents in your metro, and exactly what is free versus what needs an account.

Marketing copy reads from the scoring engine

/report and How It Works now read their dimensions and weights from the same list the scoring engine uses — the eleven categories, their point values, and which ones need an account — and a test fails the build if the two ever diverge. The free-versus-account line is stated precisely: an anonymous scan gives you the score, the ranked issues, and the full on-site breakdown; the four engine answers unlock once you have an account.

The engine roster on our pages matches what we run

Every page now names the four engines we probe — ChatGPT, Grok, Google AI, and Claude — model names included. Under the hood, the calls that turn an engine's answer into a stored fact — pulling agent names out of prose, matching identities, extracting your site's details on every score — moved to the newest Grok model available to us, because those are the calls where a wrong guess becomes a wrong number on your report.

From the blog

• Luxury Presence vs. AgentFire vs. GeoGenius: 2026 Comparison — where a website platform ends and AI visibility begins.
• The State of AI Visibility in Real Estate Right Now — what the engines do today, sourced from the study.

Two weeks of work on the front door. The blog now comes to you, the free score page was rebuilt around the question an agent actually has, and the study behind our numbers became readable by the engines it's about.

New posts land in your inbox

Every new post on the blog is now emailed the morning after it publishes. Account holders are on the list automatically and can turn it off with the "New blog posts" toggle on the account page; anyone else can subscribe from the blog index or the bottom of any post with just an email address — no account needed. Every email carries a one-click unsubscribe that doesn't ask you to log in.

Underneath is a send ledger recording who received which post, so when a send retries — and eventually one always does — nobody gets the same post twice.

/score asks the question that matters

The free score page was rebuilt from the ground up. The headline now reads "AI can't recommend an agent it doesn't know," because that's the actual problem: before an engine can name you, it has to find your site, understand what you do, and see proof you're good at it. One field above the fold, the free-versus-account split stated before you commit, and a bare domain like janesmith.com now works without the https://.

Every number on the page traces to our 36-market study.

The study is readable by the engines it's about

Our 36-market AI citation study is now served as real HTML, so GPTBot and its peers read the same 720 probes and 4,455 citations you do — a page about what AI engines cite, readable by the engines.

The findings got sharper too. The "19 universal domains" finding is reframed around what it shows: 15 of the 19 are individual agents' own websites, not platforms you can join. The 288 agents ChatGPT named across 23 markets are surfaced from its Maps lookups, the note on Claude now carries the right caveat (a smaller citation pool, so wider error bars), and a new finding covers the platforms the engines never cite.

One footer, with somewhere to follow us

Every footer now links to GeoGenius on LinkedIn and X, and every marketing page renders the same footer.

From the blog

• The Compounding Advantage: Why Early Movers Win in AI Search — why the agent who is cited first keeps getting cited.

A week of making the paid tiers real. Billing gained a grace period, the pages GeoGenius builds for you now publish by plan, and the pricing page was rewritten to match exactly what each plan includes.

A missed payment no longer costs you your tier the same day

A payment problem now opens a 14-day grace window rather than changing your plan on the spot. You keep your tier through it, the dashboard shows a banner explaining what happened and what you keep, and two emails go out: one when the problem starts and one before the window closes. Only if it's still unresolved at the end does the tier step down. Fix the card in the billing portal at any point and nothing changes.

Guides and sold pages now publish by plan

Neighborhood guides and sold-listings pages are the compounding asset — pages engines can cite that keep working after you stop paying attention.

The rule: on the free plan, your first neighborhood guide and your first sold page publish; everything after that is generated and waits. On Visibility and above, everything publishes, and the moment you upgrade the entire waiting queue goes live at once. The dashboard shows what's waiting. One thing we will never do: take a published page down because of a billing state. Deindexing takes weeks to land and weeks to undo, and AI retrieval has no "uncrawl". Pages you've earned stay up.

The pricing page describes what exists

Pricing now labels what's on the roadmap — CRM integration, the referral marketplace, the embeddable widget, and review routing to Google — as Coming soon, so every line on the page is either something you can use today or clearly marked as ahead. The page also lists the most valuable thing the free plan includes: the public AI-optimized profile page with structured data.

Your profile is in the sitemap now

Agent profile pages are now listed in our sitemap, so crawlers find them directly. And when two profiles for the same agent are merged, the retired one redirects permanently to the survivor and leaves the sitemap.

From the blog

• How to Dominate Your Market in AI Search Results — the playbook, from claiming the signals engines read to publishing what they cite.
• AI Visibility Checker: Free Instant Report for Real Estate Agents — what the free score checks and how to read it.
• Why Less Than 1% of Agents Are AI-Visible Right Now — why the field is still empty, and why that won't last.
• AI Visibility vs. SEO vs. Social Media for Real Estate Agents — where each channel earns its keep.
• From Invisible to Recommended: Real Agent AI Visibility Stories — what changed for agents who did the work.