You do it with a fixed prompt set, a baseline in Google Search Console and a tracker that covers the engines your buyers use. If you want the DIY route, read Victoria Olsina’s notes on building an AI search tracker yourself.
What you’ll need
Keep the kit small. Most of the work is thinking, not software.
- Google Search Console access for the property you want to measure, with full user permissions.
- Analytics access (GA4 or your privacy-first alternative) so you can see referral sessions.
- A list of 20 to 50 buyer prompts, written the way a UK founder or trader actually asks.
- An AI visibility tracker that covers the engines you care about.
- A spreadsheet for page-level citation logs and monthly notes.
- A focused session for the first setup, then a regular monthly review.
Here is a vendor-neutral view of where baseline tracking starts. The LLM visibility meaning behind these metrics is worth understanding before you pick a tracker. Prices move often, so check each vendor before you commit.
| Tool | What it tracks | Entry price | Engine coverage |
|---|---|---|---|
| Google Search Console Generative AI report | AI visibility on Google surfaces, mapped to URLs you already track | Free | Google surfaces only; misses ChatGPT, Perplexity and Claude |
| SE Ranking AI Results Tracker | AI Overview appearances and cited source URLs per keyword | See vendor (free trial offered) | Google AI Overviews and AI Mode; UK selectable |
| Rankscale | Prompt-level share of voice, brand versus competitor mentions | See vendor | ChatGPT, Gemini, Perplexity, AI Overviews |
| Peec AI | Brand visibility, sentiment and rival share of voice | See vendor (per-domain pricing) | Multiple AI engines |
| Otterly.AI | Brand appearance for a defined prompt set | See vendor (per-domain pricing) | ChatGPT and Perplexity |
For a wider shortlist, see Victoria Olsina’s roundup of LLM SEO tools for Web3 brands.
Takeaway: start with the free Google baseline, then pay only for the engines Search Console cannot see.
Before you start
Here is what breaks. Teams track ten brand-name prompts on one engine, see themselves everywhere, and assume they are winning.
That is false confidence. Checking whether your website is invisible to AI is a useful first test. Nobody searching “best UK crypto exchange for beginners” types your brand name first.
Decide three things before you pay for anything: which engines matter, that the location is set to the United Kingdom, and which prompts you will hold fixed for at least three months.
Takeaway: lock engines, country and prompt set on paper first. Tools come second.
Step-by-step: set up AI SEO tracking
This is the sequence Victoria Olsina follows. Each step ends with a check so you know it worked.
1. Open Search Console and record your Google baseline
Open your property and look for the Generative AI performance report. Export the URLs and figures it shows for the last three months.
Check: you have a dated export saved in your spreadsheet. If the report does not appear in your property yet, note that and use standard performance data as the interim baseline.
2. Build a prompt set from buyer questions
Write 20 to 50 prompts across three crypto themes:
- Exchange comparisons: “Which crypto exchanges are registered with the FCA?
- DeFi explainers: “How does liquid staking work?
- UK compliance: “Do I pay capital gains tax on crypto in the UK?”
Pull wording from sales calls, support tickets and community channels. Keep brand-name prompts under 20% of the set.
Check: every prompt reads like something a real buyer would type, and fewer than one in five names you.
3. Choose the engines and set the UK location
Pick engines by where your buyers ask. For most crypto and DeFi brands that means Google AI Overviews, AI Mode and ChatGPT, with Perplexity as a strong fourth.
Set the country to United Kingdom wherever the tool allows it. SE Ranking’s tracker, for example, lists the UK in its country selector.
Check: your config sheet names each engine and shows the UK location set in every tool.
4. Pick a tracker and load the prompts
Match the tool to the gap. Search Console covers Google. A second tool covers ChatGPT and Perplexity. Exploding Topics maintains a list of AI Mode tracking tools for SEO and GEO if Google’s AI Mode is your priority.
Load the exact prompts from step 2. Do not paraphrase them inside the tool.
Check: the prompt count in the tracker matches your spreadsheet.
5. Tag prompts by funnel stage and topic
Add two tags per prompt: funnel stage (awareness, consideration, decision) and topic (exchange, DeFi, compliance, wallet).
This turns a wall of results into a readable report. You will see, for example, that you own DeFi explainers but vanish at decision stage.
Check: filtering by any single tag returns a clean subset with no untagged rows.
6. Log citations at page level
Record which URL got cited, not just whether the brand was mentioned. A mention is a metric. A cited URL is a task.
Log linked citations and unlinked mentions separately. Unlinked mentions point to entity and citation work; linked ones show pages worth defending.
Check: each result row has a URL column filled in or explicitly marked “none”.
7. Benchmark against three competitors
Add three direct competitors: one bigger, one similar, one smaller. Track the same prompts for all of them.
Check: you can state your share of mentions versus each rival for every topic tag.
8. Review monthly and turn gaps into content tasks
Every month, list prompts where a competitor is cited and you are not. Turn each into a content brief, a page update or an internal linking fix. Good site structure for AI search usually solves half of these.
Check: the review ends with named tasks, owners and deadlines, not a dashboard screenshot.
Tracking without action is a hobby. Tracking that feeds a backlog is a growth system.
Troubleshooting
Symptom: Results differ on every run
- Cause: LLM answers are non-deterministic. The same prompt returns different sources on different runs.
- Fix: run each prompt several times per cycle and report the share of runs where you appear, not a single yes or no.
Symptom: Brand never appears
- Cause: thin entity pages. Engines cannot work out who you are or why you are credible.
- Fix: strengthen about, team and product pages with clear facts and author signals. Victoria Olsina’s guide to E-E-A-T for Web3 SEO covers the basics.
Symptom: Traffic from AI is not visible in analytics
- Cause: analytics groups AI referrals under generic referral or direct traffic, and many AI answers produce no click at all.
- Fix: create a custom channel group for known AI referrer domains, and treat citation logs as the primary signal.
Symptom: Competitor cited, you are not
- Cause: your content is missing or not crawlable. Client-rendered dApp pages often show bots an empty shell.
- Fix: review the cited competitor page against yours, then check rendering. Start with server-side rendering for crypto sites and Victoria Olsina’s notes on preparing a crypto site for AI agents.
Takeaway: most tracking problems are content or rendering problems wearing a data costume.
Conclusion
AI search visibility rewards teams that measure the right prompts on the right engines and act on what they find. Set the baseline, lock the prompt set, log citations by URL and review monthly.
If off-the-shelf pricing does not fit your stack, read how Victoria Olsina approached building her own AI search tracker and steal the parts that work for you.
FAQ: Frequently Asked Questions
Is Google Search Console enough?
No. Its Generative AI report gives a free baseline for Google surfaces, mapped to your URLs. It does not cover ChatGPT, Perplexity or Claude, so crypto brands need a second tool for those engines.
Which engines should a crypto site track?
Track Google AI Overviews, AI Mode and ChatGPT as the core set, then add Perplexity. Choose based on where your buyers actually research exchanges, wallets and DeFi products, and set the location to the UK.
How many prompts should I track?
Track a prompt set, meaning the fixed list of buyer questions you run through each AI engine every cycle. A practical starting range is 20 to 50 prompts (an example size, not a rule) covering exchange comparisons, DeFi explainers and UK compliance questions. Keep brand-name prompts to a small minority, and hold the set fixed so month-on-month trends stay comparable. You can check it worked when the tracker’s prompt count matches your spreadsheet.
How often should I check results?
Let the tool collect data continuously, but review monthly. Monthly reviews smooth out noise from non-deterministic answers and give you enough time to ship content changes before the next comparison.
Can I track AI Mode?
Yes. Several tools now track Google AI Mode for SEO and GEO, and SE Ranking’s AI Results Tracker covers AI Mode alongside AI Overviews. Set the UK location before loading prompts.
Do tracking tools cost a lot?
It depends on prompt volume, engines and domains. Search Console is free. Paid trackers vary, and some charge per domain, so costs scale with how many brands or sites you monitor.
What should I do with the data?
Turn every gap into a task. Where a competitor gets cited and you do not, compare the pages, fix crawlability or entity signals, and update or publish content that answers the prompt directly.
Get your brand cited in AI search
Victoria Olsina’s AI SEO and GEO service helps crypto, fintech and AI brands get ranked on Google and cited by ChatGPT, Perplexity and Google AI Overviews. Book a free strategy session to see where your brand stands in AI search today.
