Back to insights

Rebranding in the age of AI: beyond generating images

AI can generate more options. Positioning, recognizability and responsibility for publishing still require judgment.

Mojtaba RashnooBranding4 min read
A distinctive red ceramic form in front of rows of identical forms, a metaphor for brand identity amid generated sameness

When generating images and text becomes easier, redesign appears easier too: a handful of logos, tones and pictures can arrive quickly. More options do not, however, establish a reason to change a brand. Before generating them, identify what no longer works. Has the audience changed, the offer evolved, or the promised experience become inconsistent with the existing appearance? A tool can assist the work, but cannot make that business decision for us.

Find a reason for change first

In a hypothetical example, a home-products maker wants a more modern identity. Perhaps its logo is illegible at a small size, customers misunderstand the product, or the manager is simply tired of the current look. Those reasons do not imply the same scope. A technical adjustment, clearer communication and a new market position each involve different questions and different commitments.

Design Council's Double Diamond places understanding the problem before developing and testing solutions. The interpretation offered here is to distinguish observation from taste: collect concrete signs of confusion before deciding what the brand should retain or change. Design Council: discovering problems and testing solutions

Do not delegate positioning to an image generator

The American Marketing Association connects positioning with a brand's place relative to competitors and the communication of value to its audience. Our proposed application is a short brief written before generating options: whom we serve, what value we offer, which signals distinguish us and which claims we can actually support. American Marketing Association: identity and positioning

The brief needs boundaries. A brand does not have to appeal to everyone or use every tone at once. For the hypothetical home-products maker, understandable use and reassurance might matter more than a luxurious presentation. That is a choice to examine with the audience, not a universal recipe. The quality of a generated picture cannot demonstrate that the positioning choice itself is right.

Use AI to explore, not to settle the direction

Once the direction is clear, generative tools can assist early sketches, tone comparisons or composition exploration. Return each output to the brief: which meaning does it clarify, and which familiar signal does it remove? A striking presentation image may fail on small packaging, a mobile screen or a single-color print. The intended use is part of the design problem.

Reduce the options with stable criteria rather than endless additional prompts. Legibility, name recognition, appropriate tone and use in actual media are the suggested criteria here. Keep a few explainable options and record why others were rejected. Generated output is material for design, not the finished design decision. A final asset still needs deliberate judgment and professional refinement.

Keep publication responsibility clear

NIST's generative-AI profile addresses risks including inaccurate information, privacy and intellectual property. The practical response proposed here is a review route: check data sensitivity and whether the environment is approved for that use, verify textual claims, and examine asset provenance and permission. A downloadable output is not automatically ready for publication. NIST: the generative-AI risk profile

For an identity change, establish the approver, authoritative version and replacement route for older assets. If packaging, the website and support replies disagree, a visual update can create new confusion. Sometimes the sensible decision is to retain the mark and improve language or experience. AI has not removed the need for judgment; it has increased production speed and made responsible selection more visible.

Sources and further reading