AI Visibility Tools: DIY vs. Paid, Cost, Transparency

You can measure your AI visibility yourself: ask ChatGPT, Perplexity and Gemini a fixed set of buyer questions, repeat each one several times, log whether your brand is named and your pages are cited, and read your analytics and server logs for AI referrals and crawler hits. DIY costs no licence fee but a few hours of work every month, and it stops scaling once you track more than a few dozen questions, markets or engines. Buy a tool when the hours cost more than the subscription, and only one that shows its methodology and lets you export the raw answers behind every number.

GEO here means Generative Engine Optimization, the practice of getting cited in AI-generated answers (Aggarwal et al., 2023). It does not mean geodata, local rankings or map-pack tracking; those need local SEO tools, which are a different category.

What an AI visibility tool actually measures

Every AI visibility tracker does the same core job: it sends a list of prompts to one or more AI systems on a schedule, stores the answers and counts how often your brand and your domain appear, next to which competitors and with which sources. Everything else (sentiment, "share of voice", scores) is derived from those stored answers. That is good news for DIY, because the core job is something you can reproduce with a spreadsheet. It is also the reason transparency matters: if you cannot see the stored answers, you cannot check any derived number.

What no tool can measure is how often real users saw an answer that mentions you. AI providers do not publish impressions or user prompts. Any tool that reports "AI impressions" or "search volume" for prompts is modelling, not measuring, and should say so.

The DIY setup: four sources you can run yourself

1. A manual answer sample

Pick 15 to 30 questions your buyers ask, with and without your brand name. Ask each one in the engines your buyers use, in a fresh session, several times, because answers vary between runs. Log per answer: date, engine, question, run, brand named (yes/no), your domain cited (yes/no), cited URL, competitors named, other cited domains. Our guide to ChatGPT brand monitoring describes the prompt set and the scoring rubric step by step.

2. Referral traffic in analytics

OpenAI states that ChatGPT adds utm_source=chatgpt.com to links in its answers, so these visits are identifiable in GA4. Perplexity visits show up as referrals from perplexity.ai when the browser passes a referrer. Put both into a custom channel group so they do not disappear under "Referral".

3. Crawler hits in server logs

OpenAI and Perplexity document their crawlers and fetchers, including OAI-SearchBot, ChatGPT-User, PerplexityBot and Perplexity-User (OpenAI crawler documentation). A grep over your access log shows whether they reach your pages and which URLs the user-triggered fetchers read.

4. A technical readiness check

Before you measure citations, confirm that AI systems can read and quote your pages at all. The free GEO self-audit walks through robots.txt, rendering, structured data and citability with free tools. Blocked crawlers or content that only appears after JavaScript runs will keep any tracker at zero.

What DIY really costs

DIY has no licence fee, but it is not free. The honest cost is time, and it grows with three multipliers: questions, engines and runs per question. A sample of 20 questions in two engines with three runs each is 120 answers per round. Reading and coding them takes one person several hours per month, plus setup time for the spreadsheet, the analytics channel and the log queries.

DIY cost drivers and when they start to hurt

Cost driver	Manageable manually	Signal that a tool is worth it
Questions tracked	Up to a few dozen	You need hundreds, or several product lines
Engines	One or two	Buyers use four or more AI systems
Markets and languages	One	Several countries or languages, each with its own answers
Frequency	Monthly	You need weekly data to see the effect of changes
Reporting	A spreadsheet for your own team	Stakeholders or clients expect recurring reports

Scripted DIY is a middle path. You can query model APIs or run an open-source tracker such as OneGlanse on your own server. That removes the manual reading, but adds developer time, API usage billed by the model providers, hosting and maintenance. Also keep in mind that answers retrieved through an API are not guaranteed to match what users see in the consumer apps, so check the method before you trust the numbers.

What paid tools cost

Published entry prices of common monitoring tools, as listed on the providers' own pricing pages and checked on 24 August 2026:

Entry prices of AI visibility monitoring tools (checked 24 Aug 2026)

Tool	Entry price	Further tiers
Rankscale	€20/month	Credit model: Essentials €20, Pro €99 with 1,200 credits
Otterly.AI	€29/month	Lite €29, Standard €189, Premium €489
Peec AI	€70/month	Starter €70, Pro €180, Advanced €360
Profound	$99/month (billed yearly)	Second tier $399/month
Semrush AI Toolkit	Add-on to the Semrush suite	Not priced separately

Prices change; check the provider's page before deciding. Compare plans by what they limit: most cap the number of prompts, engines, markets or credits, and the tier you need depends on those multipliers, not on the entry price. GEO Tool offers a free AI visibility check for technical readiness and citability; monitoring and GEO agency services are priced on request.

Transparency criteria: what to demand before you buy

A visibility score is only as good as the answers behind it. Use these criteria in any trial:

1. Documented methodology. The vendor states how prompts are sent (consumer interface or API), from which country and language, logged in or not, and how often each prompt is repeated.
2. Raw answers. For every data point you can open the full stored answer with date, engine and cited sources. A number you cannot trace back to an answer is not evidence.
3. Repeated runs. Rates are calculated across several runs per prompt, and the tool shows how many runs a percentage is based on.
4. Per-engine reporting. ChatGPT, Perplexity, Gemini and Google AI Overviews are reported separately, not blended into one score.
5. Clear definitions. "Mention", "citation" and "share of voice" are defined, and a mention in the text is kept apart from a link in the sources.
6. Export. You can export raw answers and results as CSV or through an API, so you keep your history if you switch tools.
7. Honest gaps. Failed or incomplete runs are shown as such, not counted as "not mentioned".

A simple test during any trial: take five prompts from the tool, run them yourself on the same day, and compare. If the tool's answers and yours differ systematically, ask why before you pay. Our guide to GEO monitoring dashboards has a full scoring grid for a two-to-four-week pilot.

Our recommendation: DIY first, tool second

Start with DIY for one to two months. It forces you to define the questions that matter, shows you what AI answers say about you today, and gives you a reference to validate any tool against. Fix technical blockers first, because a tracker cannot measure citations of pages that AI crawlers cannot read. Move to a paid tool when the sample outgrows one person's time or when you need weekly, multi-market data, and choose the one that passes the transparency criteria above, not the one with the most impressive score.

Frequently asked questions

Can I track AI visibility for free?

Yes, with manual effort. A spreadsheet of repeated prompts, AI referrals in your analytics and crawler hits in your server logs cover the basics. The cost is time, which grows with the number of questions, engines and markets.

What are the best AI visibility tools?

The best tool is the one whose methodology you can check and whose limits fit your prompts, engines and markets. Look for raw answers, repeated runs, per-engine reporting and export before comparing features or prices.

How do I get more AI visibility?

Make sure AI crawlers can access your pages, answer the questions your buyers ask directly and specifically, and earn mentions on sources that AI systems cite. Measure before and after each change with the same prompt set.

Is GEO visibility the same as local or geographic visibility?

No. In this context GEO stands for Generative Engine Optimization, visibility in AI-generated answers. Local visibility in maps and local packs is measured with local SEO tools.

How many prompts should I track?

For a manual start, 15 to 30 real buyer questions, each repeated several times per round, are enough to see patterns. Expand only when you act on the results of the first set.

Are AI visibility scores reliable?

Only if you can trace them. AI answers vary between runs, so a score based on single runs or hidden answers can move without anything changing on your side. Ask how many runs a number is based on and check a sample of the raw answers.