Why does AI still use my old company name after a rebrand?

Every AI answer about your company comes from one of two places.

Training data, frozen at a cutoff

When a model answers without searching, it relies on what it learned in training. That knowledge stops at a cutoff date. If your rebrand happened after the cutoff, the model has never seen the new name.

Cutoffs are not one value per product. 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 (chat-latest). Depending on routing, ChatGPT can answer from knowledge about five and a half months apart.

Live search, pulling current pages

When a model searches, it reads pages at answer time. If the pages it finds still say the old name, the answer will too. Those pages are often not yours: directory listings, review profiles, old press coverage, partner pages.

Retrieval is discretionary. As of October 2026, ChatGPT decides per query whether to search, and Claude.ai search is a user toggle. So the same question can produce the old name one day and the new name the next.

The fix depends on the path, so diagnose first.

• Citations shown, old name in them: retrieval. Open the cited pages. One of them still uses the old name.
• Citations shown, old name not in them: the model may be mixing training knowledge with what it fetched.
• No citations: likely training data, though you can't be certain from outside.

Also separate an outdated answer from an invented one. An old name that was once true is stale data. A name your company never used is a different problem, covered in AI hallucination vs outdated answer. For ChatGPT specifically, fixing ChatGPT errors walks through the trained-versus-search split.

The free checker on this site runs up to three buyer questions and shows how engines name you, which is a quick snapshot of where you stand.

The levers with evidence behind them remove blocks. They don't add signals.

• Server-rendered text. Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. A new name that only appears after JavaScript runs may be invisible to them.
• Per-bot crawler permission. OpenAI documents that GPTBot covers training and OAI-SearchBot covers ChatGPT search, and that each setting is independent.
• Google eligibility. Google states a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode, with no additional requirements.
• A plain, quotable fact. A specific sentence on a fetchable page gives an engine something it can quote.
• Third-party mentions. When engines cite other sites, being named correctly there is associated with appearing in the answer. It's an association, not a guaranteed effect.

Schema markup, llms.txt files and AI-friendly rewriting have no credible evidence as ways to get the new name into answers. Google says its generative features need no special markup or files.

1. Write one plain sentence. For example: "NewCo is the new name of OldCo. The company renamed in March 2026." Put it on your homepage, about page and a dated announcement. Keep the old name in it, so a search for the old name finds the explanation.
2. Make it fetchable. Confirm that sentence is in the raw HTML before JavaScript runs.
3. Allow the search crawlers. Check robots.txt for OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot. Then check that your firewall or CDN doesn't return 403 to them, since robots.txt can allow a bot that the network blocks.
4. Don't hide the history. Redirect old URLs rather than leaving them broken, and think before you remove old pages. See whether to delete or noindex old pages.
5. Fix the cited third-party pages. Update directories, review profiles, marketplace listings and partner pages. Ask publishers of old news stories for a correction note where it fits.
6. Recheck by platform. Engines use different indexes, so one fixing itself says nothing about the others.

The full sequence is in the rebrand checklist for AI answers.

Training data is out of your hands this quarter. A model with a cutoff before your rebrand will keep the old name in answers given without search until the vendor ships a newer model. No vendor publishes when that will be.

Allowing GPTBot doesn't speed this up for ChatGPT search, which uses OAI-SearchBot. It may let your pages enter future training data, but no one has measured the effect. A knowledge cutoff check shows which models predate your announcement, and correcting facts past a cutoff covers what to do while you wait.

What you control is the retrieval path: what your pages say, whether crawlers can reach them, and what the cited third-party sources say. Fix those, and searched answers have the right material to draw on.