AI Visibility for Insurance Companies and Agents

Insurance questions are local, regulated and specific. Buyers ask AI to shortlist agents, explain coverage and walk them through claims. These are hypothetical prompts of the kind a buyer might type:

1. "Which independent agents in Columbus, Ohio write commercial auto for small trucking fleets?"
2. "Is [Agency name] licensed to sell homeowners insurance in Florida?"
3. "What liability limits does [Carrier]'s basic auto policy include?"
4. "Does a standard homeowners policy from [Carrier] cover flood or hail damage?"
5. "How do I file a claim with [Carrier] after a car accident, and what's the claims phone number?"
6. "Cheapest renters insurance for a student in Austin, with quotes."
7. "Which carriers does [Agency name] represent for business owners policies?"

Each prompt asks for a fact that can be checked: a state, a line of business, a limit, a phone number. That's good news. A specific fact stated clearly on a page an engine can fetch gives it a sentence it can quote. It's also the risk. When the engine can't find that sentence, it fills the gap from older training data or another site.

One more wrinkle: the prompt a buyer types is not the query the engine runs. OpenAI says ChatGPT search typically rewrites a question into one or more targeted queries. So don't chase exact phrasings. Make the underlying facts easy to find.

As of October 2026, five surfaces carry most of these questions, and each works differently.

• Google AI Overviews and AI Mode. Google says a page must be indexed and eligible to show with a snippet to appear as a supporting link, with no additional technical requirements. Organic ranking still overlaps: Ahrefs found in March 2026 that 37.9% of URLs cited in AI Overviews also appeared in the first 10 results.
• Gemini. A separate Google product, governed in part by the Google-Extended control rather than Search alone.
• ChatGPT. Search is controlled by OAI-SearchBot. The app decides per question whether to search at all. Without search, it answers from training data with a fixed cutoff, so a new state license or a changed claims number may not exist in that answer.
• Perplexity and Claude. Both run their own crawlers and retrieval, with separate robots.txt tokens.
• Microsoft Copilot. Grounds on the Bing index, so Bing Webmaster Tools health matters separately from Google.

The practical point: an agency can look fine in one engine and be absent or wrong in another. Check each one on its own. The same holds in other regulated categories, as our page on AI visibility for fintech companies explains.

AI platforms get specific facts wrong about specific businesses. For insurers and agents, the riskiest facts share three traits: they're precise, they change, and they vary by state.

• States served and licensing. An agency adds or drops a state. Old directory pages and the model's training data keep the earlier list.
• Coverage limits and inclusions. Limits differ by state, product tier and endorsement. An engine can blend a carrier's national marketing copy with one state's filing and present a limit that applies nowhere.
• Appointed carriers. Agents gain and lose appointments. Aggregator pages rarely update.
• Premiums and quotes. Engines may quote a figure from an old article or a competitor. Our guide on when ChatGPT gets your pricing wrong covers this in depth.
• Claims contacts and process. A wrong claims number is the most damaging error here, because the buyer acts on it immediately.
• Whether you still operate. Mergers, rebrands and closed branch offices can lead an engine to say the business is gone. See why AI might say your company closed.

Before fixing anything, work out whether an error is invented or simply out of date. The fix differs, and this comparison of hallucinated and outdated answers shows how to tell them apart.

Many insurance sites put their most useful facts inside an interactive quote tool: the states you write in, the lines available, the coverage options. Those tools usually load in the browser through JavaScript.

Vercel's study of AI crawler traffic found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. Googlebot does render JavaScript, so Google surfaces are less affected. Most other AI crawlers see only the raw HTML.

What this means: if "We write homeowners and auto in Ohio, Indiana and Kentucky" only appears after a ZIP code is entered, most AI crawlers never see it. Fixing this doesn't make you more likely to be chosen. It decides whether your facts exist for the engine at all.

The next move:

• Publish a plain-HTML page listing every state you're licensed in, every line you write, and the carriers you represent.
• State coverage basics in sentences on product pages, with the state they apply to.
• Put the claims phone number and process in server-rendered text, not in a chat widget.
• Keep the quote tool. Just make sure nothing important lives only inside it.

To test a page, load it with JavaScript disabled. Whatever disappears is what most AI crawlers miss.

Access is where the strongest evidence sits. These checks remove blocks. None of them adds a ranking boost.

robots.txt, per bot

Rules apply bot by bot. OpenAI states each setting is independent: GPTBot governs training use, OAI-SearchBot governs ChatGPT search. A blanket AI block often removes search eligibility by accident. A setup that allows search while declining training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

Firewall and CDN

OpenAI's help page says eligibility requires allowing OAI-SearchBot and confirming your host or CDN allows traffic from its published IP addresses. A firewall can return 403 to a crawler that robots.txt allows, and nothing in your analytics will show it. Cloudflare also changed its AI crawler defaults for some customers from 15 September 2026, so a site can lose access without anyone editing it. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.

Snippet controls

noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers on Google and Bing. Compliance teams sometimes apply them to disclosure blocks, and templates spread them further than intended. Check that your coverage and licensing text isn't tagged.

When engines cite other sites instead of yours, being named accurately on those sites is the route into the answer. That's an association in the evidence, not a promised effect. For insurance, check:

• Your state insurance department's license lookup. Engines and buyers treat it as the record. Make sure your legal name and DBA match your site.
• Carrier "find an agent" pages. Confirm each carrier that appoints you lists you, with current states and lines.
• Your Google Business Profile and Bing Places listing. Office addresses, hours and phone numbers feed local answers.
• Review and directory sites that show up when you ask engines about your agency. Note which ones they cite.
• Comparison and quote aggregators. These often carry stale premiums and carrier lists.

Ask an engine about your agency, read the cited links, and fix the ones that are wrong at the source. Request edits from site owners. You can't directly edit a model's answer, though you can report errors, as covered in whether you can ask OpenAI or Google to correct your information.

Say you run a hypothetical three-office independent agency licensed in two states. A realistic month looks like this.

Days 1 to 7: baseline. Write eight buyer questions like those above. Ask each engine, logged out, and record what it says about your states, lines, carriers and claims contact. The free checker on this site gives a quick snapshot for three buyer questions. You can also check for free whether ChatGPT mentions your brand.

Days 8 to 14: access. Audit robots.txt per bot. Test firewall responses to each AI crawler. Remove accidental nosnippet or noindex tags. Confirm index status separately in Google Search Console and Bing Webmaster Tools.

Days 15 to 21: plain-HTML facts. Publish or fix the licensing, lines and carriers page. Move coverage basics and the claims process into server-rendered text. State one fact per sentence, with the state it applies to. Make sure your own pages don't contradict each other.

Days 22 to 30: third parties and re-check. Correct carrier listings, business profiles and directories. Rerun the same questions and compare with the baseline. Expect slow movement: an engine's training data changes only with a new model, and retrieval varies by query.