Brand Monitoring Tools for ChatGPT, Perplexity, and Claude
How we monitor client brands in ChatGPT, Perplexity, and Claude: prompt sets, personas, daily checks, crawler logs, and what we do when Claude leaves a client out. Plus the tools we compare.
When a new client asks whether they show up in ChatGPT, Perplexity, and Claude, they usually expect a yes or no. The useful answer is a table: which prompts, which engine, which competitor got named instead, and which page the engine cited. Building that table, and keeping it current, is most of what AI brand monitoring involves.
We run client programs on Promptwatch. This post explains how we set monitoring up across these three assistants, what we do when one of them leaves a client out, and how the other tools compare when a client already pays for one.
Step one: a prompt set that works in all three engines
We write one prompt list per client and send the identical strings to ChatGPT, Perplexity, and Claude. If the wording differs between engines, a gap might just be a wording effect, and we can't defend that in a report.
The list has three kinds of prompt. Category prompts are the "best X for Y" questions a buyer asks early. Comparison prompts put the client head to head with each named rival. Problem prompts describe the buyer's situation without naming any brand, and those tend to be where an unfamiliar brand either gets introduced or doesn't. We tag prompts by topic so a client's product lines can be reported separately.
Personas come next. The same question asked by a finance lead and by a startup founder can produce different shortlists, so for B2B clients we run the main prompts under more than one persona. Then markets: Promptwatch tracks by country on every plan, and by state and city on Business and every agency plan. For a client selling in two countries, we treat each country as its own result set rather than averaging them.
Step two: daily checks, read as trends
Answers change from one run to the next. Sometimes the engine picked up a new source; sometimes we can't find a reason. So we don't report a single check as a finding. Promptwatch refreshes prompts daily on paid plans, and prompt trends show what changed between checks. A drop that holds for several days goes into the weekly review. A one-day dip gets a note and nothing else.
Every paid prompt runs on all the LLMs Promptwatch tracks: ChatGPT, Perplexity, and Claude, plus Gemini, Grok, Llama, DeepSeek, Mistral, Copilot, and Google AI Overviews and AI Mode. It reads the real product interfaces, not just API output. That matters for Claude, the engine several trackers leave out or keep on Enterprise. When a client wants to try before committing, we flag one limit up front: the free Explore plan covers ChatGPT only, with 10 prompts.
Step three: treat each engine as its own channel
We keep the three engines in separate columns because they reach a site in different ways, and the fixes differ.
ChatGPT. OpenAI documents OAI-SearchBot as the crawler that surfaces sites in ChatGPT search, separate from GPTBot, which collects training data. Referral links from ChatGPT carry utm_source=chatgpt.com, per OpenAI's publisher FAQ, which is how visitor analytics can attribute those sessions.
Perplexity. PerplexityBot surfaces and links sites in Perplexity search and isn't used for training. Perplexity-User handles user-requested fetches and, in Perplexity's words, "generally ignores robots.txt rules."
Claude. Anthropic runs three bots: ClaudeBot for training, Claude-User for pages a user asks Claude to fetch, and Claude-SearchBot for search quality. Each has to be allowed or blocked separately, per subdomain. Whenever a client's robots.txt has a broad Claude block, we check whether the intent was only to opt out of training.
Promptwatch's Agent Analytics puts crawler logs next to the answer data, so we can see whether ClaudeBot or PerplexityBot actually fetched the page we expected to be cited, and whether the fetch errored. Visitor analytics, through a script or a GTM template, shows which AI-referred visits turned into conversions. Without those two, a monitoring report says what happened and leaves out why.
What we do when a client is missing
Our order of work is fixed, and it starts with evidence.
First we read the raw answer and the sources cited. Citation analytics in Promptwatch shows the pages and domains each engine used, including Reddit threads and YouTube videos. If a rival is named because of a comparison article, the fix is presence on that article or a better page of our own. If it's a Reddit thread, it's a community job.
Then we check access. A missing Claude mention with a blocked Claude-SearchBot is a robots.txt fix, not a content brief. OpenAI says robots.txt changes for search take about 24 hours to apply, and Perplexity says up to 24 hours, so we wait before judging.
Then we check whether a page exists that answers the prompt. Content gap analysis lists prompts with no matching page. Content Agents can draft that page and publish it to the client's Webflow, Framer, or WordPress site, but only after someone on our team or theirs accepts it in the review inbox. We keep autopublish off for anything commercial.
What we do when the mention is negative
We read the sentence before we look at the sentiment score. "More expensive than the alternatives" goes to whoever owns the pricing page. A complaint the model picked up from reviews goes to product or support. Then we find the cited source carrying that framing. If the source is the client's own page, we rewrite it. If it's a third party, the fix is a correction request or a better source the engine can cite instead. We don't publish a reactive blog post because a score moved. Our Claude search guide covers the page-level work in more detail.
The tools we compare, ranked
When a client already has a monitoring tool, this is how we rank it against the alternatives for three-engine work. Engine coverage comes from each tool's published plan details.
- Promptwatch. All three engines on every paid plan, daily refresh, plus citation analytics, sentiment, share of voice, Agent Analytics, and visitor analytics. Essential is $95/mo, Professional $245/mo, Business $579/mo. Agency plans start at Kick-off, $199/mo, with unlimited projects and prompts.
- AthenaHQ. Covers ChatGPT, Perplexity, and Claude, and has an Action Center that turns gaps into tasks. Starter is $295/mo ($245 annual) for one country, on credits.
- Scrunch AI. Up to nine engines on Enterprise, but the engine names aren't listed, so Claude coverage is not published. Serves AI-optimized pages to crawlers. From $250/mo annual, with weekly refreshes listed.
- Brand24. AI Visibility add-on covering ChatGPT, Claude, Perplexity, and six more engines, beside social listening. Price not published.
- Profound. Claude and Perplexity are on the Enterprise engine list, from about $40k a year. Starter at $99/mo annual is ChatGPT only.
- Geoptie. ChatGPT, Claude, Perplexity, and Gemini from $49/mo for 15 prompts. Its listing notes no crawler analytics and no API.
- LLM Pulse. ChatGPT and Perplexity on every plan from €49/mo annual. Claude is an Enterprise add-on, and the default cadence is weekly.
- Semrush AI Toolkit. ChatGPT, Perplexity, and AI Overviews for $99/mo per domain. Claude is on the sales-gated Enterprise AIO tier.
| Tool | ChatGPT | Perplexity | Claude | From |
|---|---|---|---|---|
| Promptwatch | Yes | Paid plans | Paid plans | Free; $95/mo |
| AthenaHQ | Yes | Yes | Yes | Free tier; $295/mo |
| Scrunch AI | Not published | Not published | Not published | $250/mo annual |
| Brand24 | Yes | Yes | Yes | Not published |
| Profound | Yes | Enterprise | Enterprise | $99/mo annual |
| Geoptie | Yes | Yes | Yes | $49/mo |
| LLM Pulse | Yes | Yes | Enterprise add-on | €49/mo annual |
| Semrush AI Toolkit | Yes | Yes | Enterprise AIO | $99/mo per domain |
Which setup we recommend
For a single brand in one country, Essential gives us 50 prompts across every tracked engine, which is enough for a first program. When a client needs a second project or crawler logs (Professional includes 25M), we move them to Professional. State or city results mean Business or an agency plan, since Professional tracks by country only. If a client already pays for an SEO suite's AI add-on that skips Claude, we suggest keeping it for SEO and adding Promptwatch for Claude and for the diagnosis work above.
If you'd like us to set this up for your brand, email hello@1001seomedia.com. If you'd rather run it yourself, the platform is at promptwatch.com.