GO Car Rental Georgia case study: rebuilt pages, booking assistant and search growth
Case Study

Rebuilding GO Car Rental Georgia for Search, AI Assistants and Direct Bookings

6 min read | October 2026 | WordPress, SEO, booking assistant, GA4

GO Car Rental is a Tbilisi car rental company with its own fleet of 30+ cars, 24/7 handover at Tbilisi International Airport and delivery to Kutaisi Airport, Batumi and Telavi. When we started in September 2026 the site was a page-builder install that had been abandoned by its previous developer: slow, hard to edit, invisible in Google for anything except the brand name, and with no way to tell which bookings came from where. This is what we did, in the order we did it, and what the data says a month in.

In short

Page builder removed, 78 URLs rebuilt as clean HTML with structured data, a scripted booking assistant that mirrors the real booking rules, Google Tag Manager with a booking key event and first-touch attribution, and a weekly loop of small SEO bets scored on Search Console clicks. The first bookings from the new site arrived within three weeks, and AI assistants turned out to be the largest referrer.

1. Off the page builder, onto clean HTML

The old site ran on Elementor with a dozen plugins stacked on top. Every page was a few hundred kilobytes of nested divs, and every edit risked breaking something else. We replaced the theme chrome (header, drawer, footer, language switcher, call-now button) with one PHP snippet, and rebuilt every page as hand-written HTML generated from Python templates we keep in the project. The generators share one set of helpers: hero, search form, car cards, fee tables, FAQ sections with FAQPage schema, breadcrumbs, and one-way route blocks. A page is built, verified against WordPress’s own paragraph filter so nothing gets mangled on save, and pushed through the REST API.

The result is a site where the Tbilisi Airport pickup page, the Kutaisi and Batumi pages, the one-way routes, the prices page and a set of driving guides all look and behave the same, load fast in dark mode by default, and can be changed site-wide with a search-and-replace script instead of a week of clicking.

2. Content built around what people actually search for

Search Console data drives every content decision. We read the query list with the Pages and Countries breakdowns before touching anything, because the raw numbers mislead: one “Gori” cluster turned out to be searches for Goris in Armenia, and the “car sharing” cluster came entirely from inside Georgia. Each content round is one bet on one query cluster, sized to at least ten clicks a month, scored 28 days after shipping on that page’s own clicks. Misses are fine; reckless is not.

The first bet was the Gudauri and Kazbegi page for the ski season: 2,000 words, a step-by-step winter section, a distances table, ten FAQs, and 24 inbound links from other pages with rotating anchors. Alongside the bets, a travel blog publishes destination and road-trip guides that link back to the car and location pages, every post checked for Yoast green before it goes live.

Hard rules in the copy
Site copy never contradicts the real fees. Pickup and drop-off fees are shown separately from the daily rate. No invented age rules, no “hidden fees” claims, one deposit rule everywhere. The generators enforce the vocabulary.
Structured data
AutoRental organisation schema, Product and Offer on every car page, FAQPage on location and guide pages, breadcrumbs, and a live Google rating pulled through the Places API every two days.
Indexing
Every new or changed URL goes to Search Console, IndexNow and Bing Webmaster Tools the same day. Ahrefs Site Audit runs on a schedule and anything that drops the health score is fixed in the next round.

3. Leo, a booking assistant that knows the real rules

Visitors arriving from a chat assistant expect to be able to ask questions. We built Leo, a scripted assistant in the site’s own voice: it collects pickup point, drop-off point and dates through predefined answers, helps choose between Economy, Comfort and Premium down to a specific car, checks live availability against the booking system, calculates the deposit, and hands over to WhatsApp with the conversation details when a human is needed. It mirrors the booking engine’s real constraints (24-hour days with a grace period, minimum rental length, fare tables) so it never promises something the checkout cannot deliver.

4. Measurement before marketing

Before any traffic work we set up Google Tag Manager with GA4 and a booking_placed key event fired on the confirmation page. The first weeks showed a problem every booking site has: a chunk of bookings came back as “Direct” because the visitor returned from PayPal. We added first-touch attribution with a cookie set on the first visit and custom dimensions for original source, medium and landing page, so a booking is credited to the channel that actually brought the customer.

The surprise in the data: ChatGPT was the single largest referrer, around 40% of sessions in the first weeks, with engagement well above search traffic. That changed the priorities. Clear, consistent, crawlable answers to the questions an assistant gets asked (fees, deposit, airport handover, licence requirements, winter driving) now matter as much as rankings, and the same pages serve both.

5. What is next

1
Authority
The site has the content but almost no links. Listings, tourism boards and travel bloggers are the next rounds, one at a time.
2
Own booking plugin
A private WordPress plugin, Go Car Rentals, is already running in test mode alongside the current booking engine with the same data, and will take over bookings once it has run clean on real orders.
3
Scoring the bets
Each content bet is scored on its own page’s Search Console clicks 28 days after shipping, and seasonal bets again at the peak.

Have a booking or service site that Google ignores?

We rebuild it on clean HTML, wire up the measurement first, and run the SEO as a series of small, scored bets.

Talk to Syntaxa →

Engineering Insights

Latest from Syntaxa Studio.

Loading latest posts