Original data · October 13, 2026
Can AI assistants read
your CPA firm’s website?
We took 200 US CPA firm websites from Google Maps in 10 metros and checked them the way AI crawlers see them. No robots.txt in the sample blocks ChatGPT’s search crawlers. One in eight blocks Perplexity’s, and all of those sites run on the same website platform. To see what an assistant reads before it names a firm, we also collected the sources Google’s AI cites in 20 cities.
Free live session on October 22: how to get recommended by ChatGPT
For US accounting firms and real estate agents. Leave your email for a seat, and before the session I run this check and an AI answer check on your firm.
One email with the date and link, one reminder. No newsletter.
Tim Muenzel
builds and runs the systems himself
How we checked
On October 13, 2026 we pulled CPA firms with a website from Google Maps in Austin, Dallas, Houston, Miami, Atlanta, Charlotte, Phoenix, Denver, Nashville and Seattle: 200 unique websites, 184 of them reachable. For each one we read robots.txt with a standard parser for seven crawlers (OAI-SearchBot, ChatGPT-User, GPTBot, PerplexityBot, ClaudeBot, bingbot and Googlebot), loaded the home page without running JavaScript and counted the visible words, and looked for structured data that describes the firm. We loaded pages with almost no text again in a real browser to see what JavaScript adds. On October 14 we read the 184 robots.txt files again with a second parser that follows the robots.txt standard (RFC 9309), with the same counts for all seven crawlers.
ChatGPT is allowed in everywhere
None of the 184 sites blocked OAI-SearchBot or ChatGPT-User. OpenAI documents OAI-SearchBot as the crawler that surfaces websites in ChatGPT’s search features, and ChatGPT-User as the one that opens pages for a user. 25 sites (13.6%) blocked GPTBot, which OpenAI uses for training its models. A firm can block GPTBot and still show up in ChatGPT search.
One in eight sites is closed to Perplexity search
24 sites (13%) disallow PerplexityBot in robots.txt. All 24 run on the same website platform for accounting firms and share one robots.txt: it lists the crawlers it allows, including Googlebot, bingbot, OAI-SearchBot, ChatGPT-User, Claude-SearchBot and Perplexity-User, and ends with a rule that blocks every crawler not on the list. PerplexityBot is not on the list. Perplexity’s documentation says PerplexityBot is designed to surface and link websites in Perplexity’s search results and is not used to train models, while Perplexity-User only fetches a page a user asks about. The same file also blocks Claude-User, which Anthropic uses when a Claude user asks it to open a page. Two lines in robots.txt fix it, and the platform has to add them or let the firm do so.
Three sites are blank without JavaScript
3 sites (1.6%) showed no text without JavaScript and 300 to 550 words once a browser ran it. Vercel and MERJ analyzed AI crawler traffic and reported on December 17, 2024 that none of the major AI crawlers rendered JavaScript. Our own server logs show that this no longer holds for every crawler: on October 14, 2026, OpenAI’s search crawler ran the JavaScript on 4 of the 7 pages it fetched from our site, while ChatGPT-User, PerplexityBot and GPTBot showed no sign of it. A crawler that reads only the HTML still gets an empty home page from these three firms, while Google renders JavaScript and sees the text. Another 65 sites (35%) had no structured data that names the firm as an organization or local business. Corrected on October 15, 2026: we first reported 82 (45%), because our script missed types written as a list, such as "@type": ["AccountingService", "LocalBusiness"], and checked only the first eight types on pages with many. Reading those 82 sites again, 17 name the firm; the one we could not read again stays as first counted.
What Google’s AI cites for “best CPA in [city]”
The same day we asked Google, from the US, “best accountant for small business in [city]” and “best CPA firm in [city]” for 20 large US cities and collected the AI Overviews. 32 answers appeared, 25 with sources. Google cited Yelp’s search page for the city in 15 of the 25 (60%) and its own business listings in 10 (40%). ClearlyRated’s city pages, Thumbtack and Reddit came up in 3 each. Two firms that publish one page per city showed up as sources in 6 and 4 different cities. The answers drew on firm websites, Yelp and curated rating lists. ChatGPT leans on Yelp even more: BrightLocal found Yelp in 80% of local ChatGPT answers, and SteadyDemand found Yelp data attached to ChatGPT’s business cards in 95.8% of runs (August 21, 2026).
What the named firms have in common
Google’s AI named 17 firms together with a rating and a review count. Their median was 85 reviews at 4.9 stars, and none had a rating below 4.3. Among all 200 firms in our Google Maps sample, the median was 30 Google reviews, and 80 of the 200 had at least 37 reviews and 4.5 stars. The named firms had close to three times as many reviews as the median firm in the sample.
Real estate agents: a different set of sources
For “best real estate agent in [city]” Google showed 17 AI Overviews, 15 with sources. It cited its own business listings in 13 (87%), RealTrends rankings of agents by sales volume in 7 (47%), and the US News real estate agent directory in 3 (20%). Zillow and Redfin came up in 2 each. The 5 agents named with a rating and a review count had a median of 1,136 reviews. For agents, the rankings in these answers count, and RealTrends builds its rankings on verified sales volume.
What to do this week
Open yourdomain.com/robots.txt. If it ends with “User-agent: *” and “Disallow: /”, check that PerplexityBot, OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, bingbot and Googlebot are named above that line, and ask your website provider to add the missing ones. Load your home page with JavaScript switched off; if it is empty, ask for server-side rendering. Then look at the sources above for your city: your Yelp profile, Google Business Profile, rating lists such as ClearlyRated, and a page on your own site for each service and city. Measure again with the same questions after 30 days.
The numbers at a glance
Measured on October 13, 2026. Websites: 200 US CPA firms from Google Maps in 10 metros, 184 reachable. AI answers: Google AI Overviews for 60 questions in 20 US cities.
| Sites that block OAI-SearchBot or ChatGPT-User | 0 of 184 (0%) |
|---|---|
| Sites that block GPTBot (OpenAI training) | 25 of 184 (13.6%) |
| Sites that block PerplexityBot (Perplexity search) | 24 of 184 (13.0%), all on one platform |
| Home page empty without JavaScript | 3 of 184 (1.6%) |
| No structured data naming the firm | 65 of 184 (35%), corrected October 15 |
| Accounting AI Overviews citing Yelp | 15 of 25 with sources (60%) |
| Reviews of firms named by Google’s AI | Median 85 (17 firms), against 30 for all 200 firms |
| Agent AI Overviews citing RealTrends | 7 of 15 with sources (47%) |
The AI answer samples are small and come from one day and one engine, so read the shares as a direction. The website checks follow fixed rules, and anyone can repeat them.
Common questions, answered straight
Does blocking GPTBot keep my firm out of ChatGPT?
No. GPTBot collects data for training OpenAI’s models. ChatGPT search uses OAI-SearchBot, and ChatGPT-User opens the pages a user asks about. If your robots.txt allows those two, ChatGPT can read your site.
Which crawlers should an accounting firm allow?
For AI search: OAI-SearchBot and ChatGPT-User for ChatGPT, PerplexityBot and Perplexity-User for Perplexity, Claude-SearchBot and Claude-User for Claude, Googlebot for Google’s AI Overviews and AI Mode, and bingbot for Bing and Copilot. Training crawlers such as GPTBot and ClaudeBot are a separate decision that does not affect search. Google-Extended does not affect Google Search, but Google says it also decides whether Gemini Apps may use your pages to ground their answers.
How do I know if my website platform blocks Perplexity?
Open yourdomain.com/robots.txt in a browser. If you see “User-agent: *” followed by “Disallow: /” and PerplexityBot is not named above it, Perplexity’s search crawler is blocked. Every site on that platform in our sample had the same file.
Can you run this check on my firm?
Yes. Every firm that registers for the free session on October 22 gets this website check and a check of what ChatGPT, Gemini, Perplexity and Google AI Mode say about it, before the session.
Sources
- OpenAI: overview of OpenAI crawlers (OAI-SearchBot, ChatGPT-User, GPTBot)
- Perplexity: Perplexity crawlers (PerplexityBot, Perplexity-User)
- Anthropic: how Anthropic crawls the web (ClaudeBot, Claude-User, Claude-SearchBot)
- Vercel: The rise of the AI crawler, 17 December 2024
- BrightLocal: AI directory sources, 60,970 checks, 26 August 2026
- SteadyDemand: ChatGPT’s local results aren’t coming from Foursquare, 21 August 2026
- Google: Google’s common crawlers (Google-Extended does not affect Search)
- IETF: RFC 9309, Robots Exclusion Protocol, September 2022
- Open data and check script for this study on GitHub (data CC BY 4.0, code MIT)
Related reading
- How to get recommended by ChatGPT: the 2026 data
- AI visibility for CPA and accounting firms
- Free live webinar on October 22: how to get recommended by ChatGPT
- Original data: can AI assistants read real estate agent websites? 197 agents checked
- The free AI visibility check
- The AI visibility audit for accounting firms: 20 questions, four assistants, $490
- GEO for accounting firms: what it is, what works and who offers it
- Free AI crawler checker: can ChatGPT, Perplexity and Claude read your website?
- Free schema markup generator for CPA firms and real estate agents
- Original data: do AI crawlers render JavaScript? What our server logs showed
Want to know what the assistants say about your firm?
Join the free live session on October 22. Every registered firm gets this check and an AI answer check before it starts.