A rebrand or name change is one of the hardest things to communicate to AI engines, because they learn at two very different speeds. Manage that gap deliberately and you’ll keep your visibility through the transition instead of disappearing for a quarter.
The two clocks: training vs retrieval
Every AI engine combines two kinds of memory, and they update on completely different schedules:
- The training clock is slow. The model’s baked-in knowledge was frozen at a cutoff date, so it still “remembers” your old name, old logo, and old positioning until the vendor ships a retrained model — often six to twelve months later.
- The retrieval clock is fast. RAG-based engines like Perplexity and Google AI Overviews pull live, indexed pages, so they can learn your new name within days of you publishing it.
The result during a rebrand is a split-brain period: retrieval engines may show the new name while the base model’s training keeps surfacing the old one, and some answers awkwardly mix both. Understanding this is the foundation — see how LLMs learn about brands for the underlying mechanics.
Bridge the old and new names explicitly
Do not simply delete the old name from the internet. The old name is the bridge that lets a model connect what it already knows to your new identity. Instead, publish explicit equivalence statements everywhere:
- A prominent “[Old Name] is now [New Name]” banner and FAQ on your site.
- A dedicated, indexable page that states the change, the date, and the reason in plain language.
- Press coverage that uses the phrasing “[New Name], formerly [Old Name].”
This phrasing matters because it gives both the retrieval index and the next training run an unambiguous link between the two entities. The news and PR playbook is how you get that equivalence into the trusted sources models cite.
Clean up entity consistency
A rebrand is fundamentally an entity-identity problem. If the old and new names point at “two different companies” in the model’s mind, you get hallucinations, split visibility, and lost recommendations. Run a full entity-consistency cleanup using the entity building guide:
- Update structured data — change
name,legalName, andalternateName(keep the old name as an alias) in your schema, per the entity schema guide. - Fix authoritative profiles — Wikipedia/Wikidata, Crunchbase, LinkedIn, G2, and industry directories. These are high-trust sources, and inconsistency here is a top cause of AI getting your brand wrong.
- Redirect, don’t delete — 301 old URLs to new equivalents so link equity and citations carry forward.
- Keep the alias alive — retain the old name as a stated former name for at least a full training cycle.
The single most common mistake is scrubbing the old name too aggressively. Until the training clock catches up, the alias is what keeps you findable.
Monitor both names across engines
Track visibility for the old and the new name simultaneously across every engine. Build parallel prompt sets — branded prompts for each name plus category prompts — and run them with multi-engine monitoring. You’re watching for:
- New-name visibility rising on retrieval engines (the early win).
- Old-name queries correctly redirecting users to the new identity.
- Answers that confuse the two or attribute your products to the wrong name.
Expect the gap to close on retrieval engines first and base-model knowledge last.
Plan around the next model update
Because the training clock only advances at retrain time, a rebrand isn’t fully “done” until the major engines ship models trained after your change. Read navigating AI model updates and treat each new model release as a checkpoint: re-run your old-name and new-name prompt sets to confirm the model has finally internalized the transition. Until then, keep the equivalence content live and the alias intact.
Frequently Asked Questions
Why does AI still use my old brand name after we rebranded?
Base models store knowledge from a training cutoff, so they keep surfacing your old name until the vendor releases a model trained after your rebrand — often six to twelve months later. Retrieval-based engines update faster, which is why you’ll see the new name appear on some engines well before others.
Should I delete all references to our old name?
No. The old name is the bridge that lets models connect their existing knowledge to your new identity. Keep it as a stated alias (“formerly [Old Name]”) in your schema and content, and 301-redirect old URLs rather than deleting them, at least until the major models have retrained.
How long does a rebrand take to fully propagate to AI engines?
Retrieval engines can reflect the new name within days of you publishing equivalence content, but full propagation depends on the slowest training clock. Plan for a transition period of roughly six to twelve months and treat each major model release as a checkpoint.
What’s the most important thing to update during a rebrand?
Entity consistency across high-trust sources — your structured data, Wikipedia/Wikidata, and major directories. If these disagree, models may treat the old and new names as separate companies, causing hallucinations and split visibility.