From Invisible to #1 on ChatGPT in 7 Days — Here's Every Fix We Made

We took a service business in New Zealand from completely invisible on ChatGPT to the #1 recommended product in its category. The fixes took 2 hours in a single sitting. The results showed 7 days later.

This business had good Google rankings. Decent organic traffic. A professional, well-designed website. Everything looked fine on paper. But when you asked ChatGPT to recommend a business in their category, they didn't exist. Their competitors did.

What follows is every fix we made, in the order we made them, along with the day-by-day scoreboard of how ChatGPT's recommendations shifted over the following week. We've anonymised the client — treat the specifics as describing a vehicle valuation company — but every step is reproducible on any small or mid-sized business site.

The before state

On the day we ran the initial scan, the diagnostic was about as bad as it gets:

• AI visibility score: 0/100.
• Discovered in 0 of 8 target queries. Eight phrases customers actually type into ChatGPT when looking for this category of service. The business appeared in none of them.
• ChatGPT had no idea the business existed. Prompted directly by name, ChatGPT would confirm it didn't have information on the company.
• Multiple competitors were being recommended instead. Two direct competitors and one adjacent-category player were owning the recommendations. The competitor leaderboard was locked.

The site itself was well-built and ranking fine on Google. This wasn't an SEO problem. The pages read beautifully to humans and said almost nothing concrete to an AI agent. If you're used to diagnosing sites this way, you'll recognise the pattern — good Google performance and zero AI visibility is one of the most common shapes we see. We've covered the why in our post on GEO vs SEO — ranking on Google and being recommended by ChatGPT are no longer the same thing.

What we did — in a single 2-hour session

This wasn't a multi-week project. It was an afternoon. Six fixes, shipped in sequence, in one continuous work session. Here's the order we ran them in and what each one does.

Fix 1: Structured data overhaul

The site had zero JSON-LD schema markup. Nothing. No LocalBusiness, no Service entries, no Organization block — nothing for an AI agent to read. We added LocalBusiness schema with explicit service descriptions, area served, business category, opening hours, contact information, and individual Service entries for each of the main offerings.

Without getting into the code, the schema says in plain, machine-readable terms: here is the business name, here is what they do, here is where they operate, here is who they serve, here are the specific services they offer. No adjectives. No marketing. Just the facts an AI agent needs to answer a question like “recommend a vehicle valuation service in X.”

This single fix gives AI agents a machine-readable summary of exactly what the business does. It's the highest-leverage technical change on the list and, in this engagement, the single biggest driver of visibility movement.

Fix 2: Meta descriptions rewritten

Every page had vague marketing copy in the meta descriptions. Taglines and adjectives rather than facts. We rewrote them all to clearly state what the page is about and what the business does.

Before:

“Your trusted partner for all your needs.”

After:

“Independent vehicle valuation service in New Zealand. Pre-sale, insurance, and finance valuations for cars, trucks, and commercial vehicles. Reports delivered within 48 hours nationwide.”

Same length. Dramatically different signal. Meta descriptions are one of the first things AI agents read when they crawl a page — wasting them on brand copy is a silent own-goal most businesses don't know they're scoring.

Fix 3: FAQ schema added

We added FAQ pages with questions phrased exactly how customers ask ChatGPT. Not keyword-optimised questions — conversational questions, in the specific way a person would type them into an AI assistant. Then we wrapped the Q&A pairs in FAQ schema so AI agents could extract them directly.

The questions looked like this:

• “Who can give me an independent valuation on my car before I sell it?”
• “How do I get an insurance valuation for a vehicle in New Zealand?”
• “Can I get a vehicle appraisal without taking it to a dealer?”

Each answer was 60–120 words, named the business, stated exactly what the service was, and gave the specifics an AI agent would look for: service area, turnaround time, who it's for. FAQ schema has outsized effect because the content is already shaped the way AI agents consume information — the translation cost from your page to the AI's answer is almost zero.

Fix 4: Content clarity pass

We went through every key service page and rewrote the opening paragraph to explicitly state what the business does, who it serves, and where it operates. The pages previously opened with brand positioning — we reordered them so the factual block led and the positioning copy sat below it.

The rewrites removed vague language and replaced it with specific, parseable descriptions. “Serving customers nationwide” became “operating across all 16 regions of New Zealand.” “Fast turnaround” became “reports delivered within 48 hours.” “Wide range of vehicles” became “cars, light commercial vehicles, trucks, motorhomes, and motorcycles.”

Humans still get the brand further down the page. AI agents get the facts first — which is the order they need them in to answer a direct question.

Fix 5: Third-party presence

We ensured the business had consistent, complete profiles on Google Business Profile, relevant industry directories, and the key platforms AI agents cross-reference. The Google Business Profile had several empty fields — we completed the description, service categories, service area, hours, and attributes. We fixed two directory listings where the business name and phone number were inconsistent with the site. We created profiles on two category-specific platforms the business wasn't listed on but competitors were.

The point of this step is cross-referencing. AI agents are more confident recommending a business that shows up consistently across its own site, its Google Business Profile, and a handful of external directories than one that only exists on its own URL. Closing that gap wasn't creative work — it was clerical — but it was load-bearing.

Fix 6: Hosting and accessibility check

Finally, we verified that AI crawlers could actually access the site. No bot-blocking at the hosting level, no CAPTCHA walls, a clean robots.txt with no user-agent blocks against GPTBot, PerplexityBot, ClaudeBot, or Google-Extended. This site was clean — some sites aren't. If AI agents can't read your site, nothing else matters.

Total time across all six fixes: roughly two hours, start to finish. No design changes. No redesign. No new content written from scratch. Just the existing site, made readable and understandable to AI agents.

Then we waited

After the two-hour session, everything was live. We set up daily monitoring through AgenticLens to track when the changes took effect — same eight target queries, same scoring rubric, run every 24 hours. Then we waited and watched the scoreboard.

The results — day by day

Day 1–2: No change. Still invisible. 0/8 queries. Visibility score unchanged. This is normal — AI agents don't refresh their knowledge instantly, and re-crawl cycles take a day or two to complete. Most businesses quit here. You shouldn't.

Day 3–4: First appearance. The business started showing up in 2 of 8 queries. Not in the top spot yet — mentioned alongside competitors as a valid option. The structured data was being picked up and the schema was doing its job. The competitor leaderboard was still intact, but the door had opened.

Day 5–6: Momentum. Now appearing in 5 of 8 queries. The FAQ schema was landing — several of the queries where the business was now being recommended were phrased the way the FAQ entries were written. Competitors started dropping from the top position on queries they had previously owned. The business was listed second on two of the most valuable queries.

Day 7: #1 recommended for the primary target phrase. Appearing in the majority of queries. For the single highest-value query — the phrase the client had identified as their primary commercial target — the business was now the top recommendation on ChatGPT. The competitor who had dominated that position on day one was now listed second.

Seven days from the end of a two-hour work session to being the #1 recommended business on ChatGPT for the phrase that matters most. No new content. No paid placements. No product changes. Just the fixes above, shipped once, left to land.

What made the biggest difference

The monitoring data made it possible to rank the fixes by impact. Here's what moved the needle, in order:

1. Structured data was the single biggest needle-mover. Adding LocalBusiness JSON-LD with explicit, complete fields took the site from invisible to parseable. AI agents don't guess at what a business does — they read signals. Explicit beats implicit every time, and structured data is the most explicit signal available.

2. FAQ schema had an outsized effect. Because AI agents are answering questions, content that's already structured as question-and-answer is unusually easy for them to use. The queries where the business ranked earliest and highest were the ones that most closely matched the FAQ entries — the correlation was obvious in the data.

3. Content clarity made the existing pages parseable. The content was always good. It just wasn't written for AI consumption. Leading each page with a one-paragraph factual summary gave AI agents something to extract directly.

4. Third-party presence provided the confirming signals. AI agents cross-reference. Once the on-site signals were in place, the consistent off-site footprint turned a maybe-real-business into a confidently-recommendable one.

5. Meta descriptions and accessibility were foundational. Necessary but not sufficient on their own. If either had been broken, none of the other fixes would have landed. Because both were clean, they acted as a multiplier on everything else.

The takeaway

This wasn't a redesign. The site didn't change visually. The content didn't fundamentally change. The SEO didn't change. We simply made the existing site readable and understandable to AI agents.

2 hours of work. 7 days of waiting. #1 on ChatGPT.

Most businesses are sitting on a website that could perform like this with the same fixes. They just don't know what AI agents need because nobody's told them yet. If you want the longer playbook with every variation and before/after example, our guide on how to get your business to show up in ChatGPT covers the full fix list in detail.

If you run an agency and you're wondering whether this workflow is repeatable across a book of clients, it is — and the economics work. Our agency workflow is built around exactly this audit-fix-monitor cycle, with white-label reporting and per-site pricing designed for agencies running AEO as a service line.

Your next move

Want to see where your site stands? AgenticLens scans your website, runs real queries against ChatGPT, and shows you exactly what to fix. Free scan, 60 seconds, no signup at agenticlens.io.

Want us to monitor it for you? Start a free trial and watch your score climb as you implement the fixes — same daily scoring we used to track the engagement above, run automatically against your site and your target queries.

Run a free scan at agenticlens.io — you'll see your AI visibility score, which queries find you, and which ones don't. The fix list comes with it.