A recognisable brand in the age of AI: why producing more content is not enough

Until a few years ago, producing a good article, a polished carousel or ten variants of an ad took time and people. Today a company can generate hundreds of them in an afternoon, and so can all its competitors. The result is an environment full of content that is correct, tidy and alike.

The short answer is this: when production is cheap, quantity stops setting you apart. A brand is recognised through a few distinctive assets, that is, colours, shapes, photographic style, tone of voice and recurring phrases, always used together with the name and consistently across the website, social media, ads and generated content. To use AI and creative variants without losing your identity you need a written rule that says what stays fixed, what can change and what is off limits. Brand consistency, aesthetic quality and performance are three different things: they are measured in different ways and should not be confused.

Why is producing more content no longer enough?

Because AI has multiplied the supply of content, not people’s attention. The American agency Directive puts it bluntly when talking about B2B content marketing: AI has massively increased the supply of educational content, and articles that once set a company apart can now be produced in a few minutes by almost any competitor (Directive).

There is a second, less visible effect. A language model tends to produce the most probable answer, the one that most resembles what it has already seen. Mark Newton of the British agency Hallam sums it up like this: a language model doesn’t know what is true, brilliant or original, it only knows what is expected (Hallam). If everyone asks AI to “write a post about our new service”, everyone gets similar posts. The difference lies in what you put in and what you hold firm.

For buyers, recognising a brand is a shortcut. Byron Sharp of the Ehrenberg-Bass Institute calls it mental availability: the likelihood that a person notices, recognises or thinks of a brand at the moment they need to choose, and he stresses that it is much more than simple awareness (Ehrenberg-Bass Institute). A hundred pieces of content that nobody recognises as yours don’t build mental availability. Ten that are recognisable at first glance do.

What makes a brand recognisable across website, social media and ads?

Distinctive assets. Jenni Romaniuk, who has studied them for years at the Ehrenberg-Bass Institute, defines them as elements other than the name, such as colours, logos or characters, that uniquely signal the brand (Brands of Distinction). They can be:

• a colour or combination of colours;
• a shape, a graphic cut, a way of composing an image;
• a typeface;
• a photographic or illustration style;
• a character, a mascot, a recurring face;
• a sound or piece of music;
• a phrase, a way of opening or closing copy.

To assess them Romaniuk uses two criteria: fame, meaning how many people link that element to the brand, and uniqueness, meaning how far they link it only to that brand and not to competitors. Of the two, uniqueness matters more. An element that is both famous and unique can end up replacing the name in advertising; one that is famous but shared with others also calls the competition to mind. Her book Building Distinctive Brand Assets (Oxford University Press, 2018) is the reference work on the subject.

A practical detail: distinctive assets work if they are born and grow alongside the name. A colour used on its own for years, without the brand next to it, never comes to belong to anyone. And people inside the company tend to overestimate its strength: a study published in the Journal of Brand Management found that marketers overestimate how famous their distinctive assets are and underestimate how unique they are (Marketing Week).

Brand consistency, aesthetic quality and performance: what’s the difference?

They are three different judgements about the same piece of content, and confusing them leads to poor decisions.

	What it assesses	How to check it	The risk if you confuse it with the others
Brand consistency	Whether the content is recognisable as yours	Comparison with the brand rules; recognition tests without the logo	A beautiful but anonymous piece passes for a success
Aesthetic quality	Whether it is well made: composition, light, legibility	Review by someone with visual expertise	Effective content is rejected because “it isn’t pretty”
Performance	Whether it delivers the expected result in the short term	Campaign and CRM data	Identity is sacrificed for one more click

The third deserves attention. Campaign metrics measure the effect over the next few weeks, whereas recognisability is built over time. Les Binet and Peter Field, analysing the cases collected by the IPA, the UK’s agency association, identified an optimal balance of roughly 60% of budget on brand building and 40% on activation (IPA). For B2B, their later analysis, published by the B2B Institute, a think tank funded by LinkedIn, suggests roughly 46% and 54% (LinkedIn). These are averages based on large brands, not rules to apply to an SME. The point that still holds is a different one: judging branding only on a campaign’s immediate return means measuring it with the wrong tool.

Statistics such as “brand consistency increases revenue by 23%” also do the rounds. They come from surveys by a software vendor in which companies estimated how much they would earn (Inc.): they are respondents’ opinions, not measurements. Best not to build a budget on them.

The identity matrix: what stays fixed and what can change

The identity matrix is the tool that turns the brand into rules anyone, person or AI, can apply. It has five rows and three columns: what is fixed, what can vary, what is off limits.

Element	Fixed	Can vary	Off limits
Tone	Register, grammatical person, typical vocabulary	Length, technical level depending on the channel	Tones the brand would never use (irony at the customer’s expense, empty superlatives)
Visual codes	Main colours, typeface, logo position, lighting style	Subjects, framing, backgrounds, formats	Foreign palettes, trendy filters, effects that change the mood
Promise	The sentence that says what you do and for whom	The benefit highlighted for each audience	Promises the service doesn’t keep
Proof	Accepted sources: approved case studies, verified data, certifications	Which proof to use in each piece	Unverified figures, invented testimonials, client logos used without permission
Limits on variation	How many distinctive assets must always appear	The order and weight of the elements	Variants with no distinctive asset at all

To give you an idea, here is how the visual codes row reads for the covers of this blog: dark background, a single yellow, the one from the logo, used as a point of light, images in half-light with no overlaid text. Subjects and framing change from one article to the next; the yellow and the light don’t.

Illustrative image: at the top, inconsistent variants; at the bottom, different variants that remain recognisable.

The last row is the one that is usually missing. Saying “the main colour and the lighting style must always appear, plus the logo or the name” lets you generate many variants without any of them becoming anonymous.

How do you use AI and creative variants without losing your identity?

Ad platforms ask for variety. Meta explains that Andromeda, the system that selects which ads to show each person, works better with greater ad diversity (Engineering at Meta). Variety and recognisability are not in conflict, as long as the identity matrix says where you can vary. Gaetano explains how to choose the variants to test in your campaigns in what to automate in advertising campaigns.

A workflow that holds up:

1. Controlled inputs. Give the AI the identity matrix, the approved assets (logo, palette, examples of successful photos and copy) and the content brief. The more precise the input, the less the result resembles everybody else’s.
2. Generation on execution, not on identity. AI is useful for adapting formats, generating backgrounds, proposing variants of copy and framing. The concept, the promise and the proof are decided by people who know the brand.
3. Text, logos and figures kept out of the model. Lettering, logos and numbers are overlaid at the design stage, not generated by the model: it’s the simplest way to avoid mistakes and distorted logos.
4. Review against the matrix. Every variant passes the five-row check before it is published.
5. Platform tools configured. On Meta, Advantage+ creative enhancements can be switched off when they alter the image in unacceptable ways, and Meta recommends checking the preview of every ad (Meta). On Google Ads, Performance Max brand guidelines let you specify name, logos, colours and typeface, with the stated limitation that not every format applies them.

Transparency about generated content

From 2 August 2026 the transparency obligations in Article 50 of the EU AI Act apply to content generated or manipulated with artificial intelligence. Anyone who distributes deepfakes, meaning images, audio or video that realistically depict real people, places or events, must disclose it; for works that are clearly creative or fictional, the obligation is lighter (European Commission, text of Article 50). Standards such as C2PA, Content Credentials, let you attach information about origin and edits to files. For a brand, transparency is also a matter of trust: a generated image pretending to be a real photo of your factory is a risk even before it is a legal problem.

Consistency for the AI systems that talk about you, too

There is one last reason to look after consistency. AI assistants that answer questions about suppliers describe a brand on the basis of what they find. Hallam notes that consistent messaging is needed for language models to classify a brand with confidence, and that inconsistencies between channels create confusion (Hallam). It’s no guarantee of being cited, but describing yourself the same way on your website, profiles and articles helps you avoid being confused with others. We covered this in GEO: getting cited by AI and in digital PR and AI search.

How do you turn rules into a system?

A matrix left in a drawer is useless. It becomes a system when it lives inside the tools people use every day:

• templates for recurring formats (posts, carousels, ads, covers), with the fixed elements already locked;
• an asset library with clear names and approved versions, so nobody downloads the old logo from an email sent in 2021;
• base prompts for AI tools, containing the tone, visual codes and prohibitions from the matrix;
• a review grid with the five rows, to be used before every publication;
• one person in charge who decides on borderline cases and updates the rules.

The operational side for campaigns, with formats, templates, file organisation and creative testing, is in our guide to the brand kit for advertising. For consistency in your social calendar there’s the guide to the social media editorial plan.

How can you tell whether your brand is recognisable?

With a few measures, repeated over time:

• recognition tests: show a sample of your audience a piece of content without the logo or name and ask whose it is. Romaniuk warns that questions which suggest the answer inflate the results (Ehrenberg-Bass Institute), so the question must be open;
• branded searches: the trend in searches for your brand on Google and in direct traffic to your website, read over long periods;
• sales listening: what new leads say when they explain how they heard of you;
• spot checks: every month, ten published pieces compared with the matrix.

Mistakes to avoid

Measuring the brand with campaign metrics. This week’s clicks don’t tell you whether the brand has become more recognisable.

Changing style every season. Distinctive assets become yours through repetition. Changing them often wipes out the work done.

Leaving identity decisions to AI. AI is good at variation, but it gravitates towards what is already common.

Using elements that aren’t unique. A colour or style your competitors also use brings them to mind too.

Generating text, logos and figures with the model. They are overlaid at the design stage, always.

Publishing realistic generated images without saying so. Beyond the obligations of the AI Act, it’s a risk to trust.

A checklist to get started

• List the elements you currently use on your website, social media and ads, and mark which are genuinely yours and which are common in your industry.
• Choose a few distinctive assets to use always together with the name.
• Fill in the identity matrix: tone, visual codes, promise, proof, limits on variation.
• Prepare templates, an asset library and base prompts that contain the matrix.
• Set up brand guidelines in Google Ads and check Meta’s creative enhancements.
• Decide who approves, and with which review grid.
• Schedule a recognition test to repeat once or twice a year.

Want an identity that holds up even with AI?

We analyse how your brand presents itself today across website, social media and campaigns, identify the genuinely distinctive assets and build with you the identity matrix, templates and rules for using AI in content production. We work on social media and paid social campaigns, with creative produced in-house.

You can book a call with Federico or write to us via the contact form.

Frequently asked questions

What are a brand’s distinctive assets?

They are elements other than the name, such as colours, shapes, typefaces, photographic styles, characters, sounds or phrases, that people link to a single brand. According to Jenni Romaniuk of the Ehrenberg-Bass Institute, they are assessed on two criteria: fame and uniqueness.

Does AI make all brands the same?

It tends to, because models produce the most probable result. That happens when you also ask it to decide the identity. If the identity is set down in written rules and AI works on variations in execution, the result stays recognisable.

How many creative variants do campaigns need?

The platforms don’t give an official number. Meta explains that greater ad diversity helps its selection system. It pays to vary concepts and executions while always keeping the brand’s distinctive assets.

Do you have to disclose images created with AI?

From 2 August 2026, Article 50 of the AI Act requires deepfakes to be disclosed, meaning realistic content depicting people, places or events in a way that seems authentic. For borderline cases you need legal advice; in any case, it’s good practice not to pass off generated images as real photos.

Does brand consistency mean making everything the same?

No. It means some elements stay fixed and others can change within clear limits. Subjects, formats and messages vary; colours, light, tone and promise remain recognisable.