It is 6:40pm on a Monday and you've just finished training. Outside the gym you weigh it up: order delivery or duck into the supermarket opposite the train station on the way home. You open ChatGPT on your phone and start to type: "give me a few good post workout meal options that I don't have to cook and actually taste decent".

During our August 2026 audits, in response to our prompts, none of the four big AI assistants, known as large language models (LLMs), named one healthy ready-made meal brand when asked what to eat after a workout. Ask for "high-protein ready-made meals" though and all four did. The prompt and the shopper simply asked earlier in the decision-making process. They were describing their need.

Visible on the product, invisible on the need. OVI scores for a healthy pre-prepared meal brand: 4.62 when the prompt names the brand (0% of responses absent), 2.57 when it names the product (40% absent), 1.71 when it describes the need (67% absent). All four LLMs named the brand for 'best high-protein ready-made meals in Australia' and 'best meal prep delivery service in Australia'; none did for 'what should I eat after a workout for recovery' or 'healthiest convenience foods at Woolworths'.

Naming the brand is the easy part

A prompt naming the ready-made meal brand scored 4.62 out of 5 on the Obsessd Visibility Index (OVI), our measure of how visible a brand is in AI answers (method below). Naming the product scored 2.57 and describing the need only 1.71. The pattern carries over to other categories in our research. Moving from product-based to need-based prompts, the share of responses missing the brand rose from 58% to 80% in personal finance and from 50% to 75% in property. Tertiary education moved from 12% to 38%.

The AI invisibility pattern replicates across categories. Share of LLM responses that failed to mention the brand, product-based prompt versus need-based prompt: Personal Finance 58% to 80%, Property and Developer 50% to 75%, Pet Care 43% to 58%, Tertiary Education 12% to 38%.

One directory becomes the answer

A key finding from our research across multiple pet care brands. Asked for the best dog food in Australia, Claude and Gemini chose Royal Canin while ChatGPT chose Hill's Science. Perplexity chose Lyka, drawing on one review directory behind 12 of 24 pet-care answers, half of those we collected from that LLM.

Why brands lose and what is typically broken

Brands lost recommendations for six reasons: value, expert authority, disintermediation, being the baseline comparator, reputation and business model. One Australian online travel brand lost on all four LLMs. The advice: "Use Google Flights or Skyscanner to search, then book directly with the airline." Underneath, product pages often lack schema markup and JSON-LD, the structured code that tells machines what a page says. Wrong prices and ingredients can get repeated as fact.

What to do

Publish current, citable authority, because older coverage can outweigh newer facts. Favour third-party validation over brand claims. Address the head-to-head comparison directly. Track named comparison prompts as well as category rankings. Agencies can add need-based and comparison prompts to client reporting.

Customers who name a brand tend to find it. The commercial question is whether it is found when you only describe what you need. For a free audit and remediation plan, contact hello@obsessd.ai.

Research methodology

In this research we tested ChatGPT, Claude, Gemini and Perplexity. Every prompt comes from a six-category library built on accurate brand facts, tracked competitors, Search Console, Ahrefs and GA4 data, spanning use case, occasion and seasonality. Each answer is scored on the OVI out of 5. Visibility is how prominently the brand appears. Accuracy is whether what the LLM says is true. Competitive Position is whether the brand beats the rivals named in the same answer. Audits are both manual and API based for accuracy. Results reflect LLM outputs at a point in time and are not a view on product quality. Research was not commissioned or funded by any brand. Obsessd sells audit and remediation services. To request a correction or raise a concern, contact hello@obsessd.ai.