Which AI platforms should my brand track first?

Start with the engines your buyers use when they research a purchase. That is the only ranking that ties tracking to revenue. Then measure each engine on its own, because a combined score can hide an engine where you're missing entirely.

The order looks like this:

1. Find where your buyers ask. Use customer conversations, sales call notes and referral data.
2. Track those engines first, each with its own result. Our guide to measuring visibility across models covers why per-model results matter.
3. Check access for every engine, tracked or not. Access checks are cheap and the failures are invisible.

Tracking costs prompts and time, so it should be selective. Access problems cost you every engine at once, so the check should not be.

Each engine retrieves from its own index, and an engine can't cite a page its index no longer holds. OtterlyAI's 2026 experiment shows how wide the gap can get.

Otterly published about 1,000 AI-generated posts on each of two new domains. Google deindexed both, with no manual action reported. Bing kept both. Otterly tracked 3,934 citations for one site, capybara-world.com, and 2,718 of them (69%) came from Microsoft Copilot, which grounds on Bing. Perplexity and Claude also cited the sites. ChatGPT, Gemini, AI Overviews and AI Mode almost never did, and ChatGPT cited them effectively zero times even while the sites ranked in Bing.

The test has limits: one site per arm, a low-stakes topic, and a vendor running it. Its lesson for tracking still holds. One engine can cite you thousands of times while others barely notice you. If you'd tracked only one engine, or blended them, you would have drawn the wrong conclusion.

Retrieval also varies per query. As of October 2026, ChatGPT decides per question whether to search, AI Overviews often don't trigger, and Claude.ai search is a user toggle. So an engine's answer about you may come from live retrieval or from what the model learned in training.

"Google AI" is not one thing to track. Google states that AI Overviews and AI Mode "may use different models and techniques" and can return different links. The Gemini app is a third product, governed by a different crawler control (Google-Extended), and Gemini API grounding is a fourth retrieval path.

The rules for entry also differ by surface. For AI Overviews and AI Mode, Google says a page must be indexed and eligible to show with a snippet, with no additional technical requirements. There is no robots.txt token that blocks AI Overviews alone: Googlebot rules and snippet controls are the levers.

If Google is where your buyers search, decide which surface matters. AI Overviews appear inside ordinary results. AI Mode is a separate conversational experience, covered in whether you can track Google AI Mode. Track them as separate rows.

No public dataset tells you which engine your specific buyers use. You have to assemble it from your own signals.

• Ask. Add "where did you hear about us?" options that name AI assistants, and listen for them in sales calls.
• Read referral data with care. ChatGPT adds utm_source=chatgpt.com to referral links, while Google AI Overviews send no referral at all. Referral share overstates engines that tag their links. See AI referral traffic tracking for the setup.
• Use engine-reported data where it exists. Google Search Console shows AI Overviews and AI Mode impressions only. Bing Webmaster Tools reports citations for Copilot and Bing AI summaries. The trade-offs are in prompt tracking vs engine-reported data.

Say you sell software to developers. If prospects mention Perplexity and Claude in calls, those move up your list, whatever the general headlines say.

Access is the one area with strong evidence and full publisher control. Robots.txt rules are per bot, and OpenAI states each setting is independent: GPTBot covers training, while OAI-SearchBot controls ChatGPT search. Blocking or allowing one doesn't change the other.

Three more checks apply to every engine:

• Firewall and CDN. A CDN can return 403 to a crawler that robots.txt allows. Since 2026-09-15, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and free-tier customers.
• Rendering. Vercel found ChatGPT and Claude crawlers fetch JavaScript but don't execute it. Facts that only appear after JavaScript runs may be invisible to them. Googlebot renders.
• Snippet controls. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers, sometimes by accident.

99Visibility's AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away.

1. List the engines your buyers mention and rank them.
2. Run the access checks above for every engine, not just the top of your list.
3. Write buyer questions for your top engines. Mix branded and unbranded prompts so you see both reputation and discovery.
4. Record results per engine and per Google surface, never as one blended number.
5. Re-check on a schedule. Our check cadence playbook covers how often.

For a first snapshot, the free checker on this site shows whether engines name you for up to three buyer questions. No tool controls what models say. You control whether they can read you.

Should I track every AI platform from day one? Not for prompts. Tracking every engine spreads your budget across engines your buyers may never use. Track the two or three your buyers mention most, and add others as evidence appears. Access is different: check robots.txt, firewall reachability and rendering for every major engine from the start, because those checks are cheap and a blocked crawler gives no warning.

Does ranking well in Bing mean ChatGPT will cite me? Not necessarily. In Otterly's 2026 experiment, the test sites ranked in Bing and ChatGPT's crawlers kept visiting, yet ChatGPT cited them effectively zero times. Microsoft Copilot, which grounds on Bing, cited them heavily. Being indexed is necessary, but each engine makes its own retrieval and citation choices, so measure ChatGPT directly.

Is allowing GPTBot enough for ChatGPT? No. GPTBot governs whether your content may be used for training. ChatGPT search uses OAI-SearchBot, and OpenAI states the settings are independent. OpenAI also says your host or CDN must allow traffic from its published search bot IP addresses. Check both robots.txt and actual reachability for OAI-SearchBot.

Can I get one score that covers all AI platforms? You can compute one, but it hides more than it shows. Engines use different indexes, cite in different units, and retrieve per query. Otterly's data showed one engine carrying most of a site's citations while others barely cited it. Keep per-engine results as the primary view and treat any blended score as a summary, not a measurement.