Rebrand Checklist: Updating What AI Says About You

AI answers come from two places, and a rebrand hits them differently.

The first is trained (parametric) knowledge. The model learned your old name before its knowledge cutoff, and that snapshot doesn't change. OpenAI's model docs (checked August 2026) list a February 16, 2026 cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path that ChatGPT also routes to. So one product can hold two different snapshots of your company. Nothing you publish this quarter reaches a retrieval-off answer. Only a newer model, trained on newer data, can carry your new name. That is why AI still uses your old company name months after launch, and why a new model release can change brand answers overnight.

The second is retrieval. When an engine searches the web before answering, it can read your current pages. This is the channel you can influence now. It runs through four gates:

• Access: can each AI crawler fetch your new pages?
• Index: does the engine's index hold them?
• Retrieval: does the engine pull them in for a given question? (Partly up to the engine.)
• Generation: does the model use and cite them? (Up to the model.)

Retrieval fires at the engine's discretion, per query. The same question can get a fresh answer one day and a stale one the next. For the ChatGPT side specifically, see trained knowledge vs search errors.

Run these in order. Steps 1 to 4 remove blocks. Steps 5 and 6 give engines something correct to find.

1. Redirect every old URL. Map each old-domain URL to its closest new URL with a permanent (301) redirect. Keep the old domain registered and redirecting. A dead old domain leaves engines with nothing current to fetch when they look up the old name.
2. Confirm crawlers can reach the new site. robots.txt rules are per bot. OpenAI documents that GPTBot governs training only, while OAI-SearchBot controls ChatGPT search, and "each setting is independent of the others." A minimal allow block looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

3. Test the firewall, not just robots.txt. A CDN or firewall can return 403 to a crawler that robots.txt allows. New domains are the risk here: from September 15, 2026, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.
4. Check index status per engine. Google says a page must be "indexed and eligible to be shown in Google Search with a snippet" to appear as a supporting link in AI Overviews or AI Mode. Check Google Search Console and Bing Webmaster Tools separately. Bing's index backs Microsoft Copilot. Also scan new templates for stray noindex, nosnippet or data-nosnippet, which keep content out of AI answers.
5. Publish one plain old-to-new statement. Put a sentence on your About page and homepage footer that states the change as a fact. Hypothetical example: "Acme Analytics changed its name to Brightline on 3 March 2026. Products, pricing and team are unchanged." A specific, fetchable fact gives an engine a sentence it can quote.
6. Make sure that statement is in the raw HTML. Vercel found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. If your name and the old-to-new statement only appear after JavaScript runs, most AI crawlers never see them. Googlebot renders; most others don't.

Copy this into your launch plan and tick items as you verify them, not as you assign them.

On your own site

• Every old URL 301-redirects to a live new URL (spot-check your top pages and the pricing page).
• Old domain stays registered and redirecting.
• robots.txt allows OAI-SearchBot, PerplexityBot, Claude-SearchBot and Meta-WebIndexer, and you've decided GPTBot, ClaudeBot and Google-Extended (training bots) on purpose.
• Firewall and CDN return 200, not 403 or a challenge page, to each AI crawler.
• No accidental noindex, nosnippet, max-snippet:0 or data-nosnippet on key pages.
• New pages indexed in Google Search Console and Bing Webmaster Tools.
• Old-to-new statement present on About, homepage and press page, in server-rendered HTML.
• Name, product names and pricing stated the same way on every page (no leftover old name in footers, page titles or legal text you didn't mean to keep).

Off your own site

• Company profiles updated: LinkedIn, app marketplaces, review sites, directories, partner pages.
• Wikidata entry updated if one exists (read whether Wikidata affects AI answers before treating it as a fix).
• Press release or announcement that names both the old and new name.
• Request updates on third-party articles and listings that rank for your old name.

The off-site list matters because when engines cite other sites instead of yours, being named correctly on those sites is the route into the answer. The evidence shows an association, not a guaranteed effect.

A log turns "AI still gets it wrong" into a list of specific, checkable problems. Start it before launch so you have a baseline.

For each test, record:

• Date, platform and model (where the app shows it).
• The question, written the way a buyer would ask it, plus one that asks for your old name directly.
• Retrieval state: did the answer show citations? If not, assume trained knowledge.
• What it said: old name, new name, both, or a merged company.
• Error type: outdated (true before the rebrand) or invented (never true). The difference between hallucinated and outdated answers decides the fix. Outdated answers from trained knowledge need time. Outdated answers with citations point to a stale page you can find.
• Cited URLs, so you can trace which source carries the old name.

Test logged out with memory off. Saved memories can shape how ChatGPT rewrites a query, so your own account isn't a clean test (see how ChatGPT's memory affects brand answers). For a broader method, follow how to find out what AI says about your company.

Good

• Answers that show citations use the new name and link to your new domain or to updated profiles.
• Answers without citations say something like "formerly known as" or use the old name for the period before the change. That's the cutoff talking, and it's expected.
• Errors in your log shrink from "wrong source cited" to "no source, old snapshot" only.
• Each engine's index shows your new pages.

Bad

• A cited answer still links to an old-domain page that returns 404. The redirect is missing.
• An engine with search on never cites your new site. Check access and index status for that engine first.
• The answer merges you with a different company that shares your new name. Make the old-to-new statement more specific (location, product, founding year).
• One engine updates and another doesn't. That's normal. Engines use different indexes, so diagnose them one at a time.

Don't read one blended number. A single score averages answers given with and without live search, which hides exactly the split you need to see.

• Launch week: run the full checklist. Re-test access and redirects after every deploy that touches routing, CDN or robots.txt.
• Weeks 2 to 8: check index status weekly and add entries to the error log. This is when cited answers should start switching.
• Monthly after that: rerun the same questions on each platform and compare with your baseline. The free checker on this site gives a quick snapshot for up to three buyer questions.
• On every major model release: rerun the log. A new model may carry your new name in its trained knowledge for the first time, or may not if its cutoff predates the rebrand.
• Twice a year: re-audit third-party profiles. Old listings resurface as sites get copied and scraped.

Stop treating the rebrand as open once cited answers use the new name across platforms and the only remaining errors are uncited, pre-cutoff answers.

Be honest with your leadership team about the limits.

• You can't update a trained model. Retrieval-off answers change only when vendors ship models with later cutoffs.
• You can't make an engine search. Retrieval is per query and at the engine's discretion.
• You can't choose which source an engine cites once it has retrieved several.

What you control is access, index eligibility, consistent facts and correct profiles. None of it guarantees an outcome. All of it removes reasons for an engine to keep repeating your old name.

Skip the shortcuts. Google's AI optimization guide says you don't need new machine-readable files, special markup or AI-specific writing for its generative features, and schema isn't a lever for AI answers. Hidden text or cloaked pages aimed at AI crawlers are manipulation, and they risk misleading answers and penalties.