Much of the work is familiar: clear product pages, accessible websites, reviews, comparisons and coverage elsewhere. Google says established SEO practices apply to AI Overviews and AI Mode, with no special markup or additional technical requirements. Google Search Central
OpenAI similarly describes product information, reviews and third-party content among the inputs that can shape shopping results. OpenAI Help Center
Updating a website or securing press coverage can be useful. Calling it AI optimisation does not, by itself, make the method new or more effective.
Third-party coverage still matters. Research published by AI-visibility platform Peec AI, based on nearly 200,000 AI responses, found that higher positions in frequently cited comparison lists were associated with better brand visibility. Additional placements showed diminishing returns, and results varied between markets and systems. These were associations, not proof that securing a placement causes more recommendations. Peec AI
This raises a familiar concern. Businesses with the resources to keep publishing and securing coverage may shape more of the material an assistant encounters. Promotional claims could then reappear as an apparently independent recommendation. That is a risk worth examining, rather than evidence that the biggest marketing budget always wins.
The more interesting difference is how the customer’s need develops.
Someone asks for a laptop, then explains their budget, frequent travel and dislike of fan noise. Each detail changes what would make a suitable recommendation. ChatGPT can use the conversation and relevant saved preferences to shape product selection and searches. The recommendation is being made against a developing situation, not just a category name. OpenAI Help Center
Assistants can also turn a question into several searches, then search again after reading the results. Google calls this query fan-out; OpenAI says ChatGPT may rewrite a request into one or more targeted queries and refine them after reviewing initial results. An answer may therefore draw on information the customer never explicitly asked to find. Google Search Central · OpenAI Help Center
That suggests more useful tests than counting mentions. Does the assistant find a business without being given its name? Does it describe the offer correctly? Does the recommendation remain suitable as the customer adds constraints? Being praised when named does not establish that a business would have entered consideration unaided.
The questions used for testing matter just as much. Adding “which product should I buy?” to an informational question steers the answer towards a purchase. It cannot establish that the original need would naturally lead there. Nor does testing thousands of invented questions reveal how often real customers ask them.
Some findings expose less obvious problems. In controlled tests with fictional companies and web search disabled, changing the company name altered how several models described otherwise comparable businesses. Occasionally the description followed the meaning of the name rather than the supplied facts. This is evidence of a weakness in interpretation, not a demonstrated strategy for improving sales. Peec AI
Discovery can now lead directly to action. Supported shopping and reservation services can connect recommendations to purchases or bookings. Here, discoverability includes whether the assistant can help complete the task, rather than simply describe a possible supplier. Availability depends on the platform and service. OpenAI Help Center
These differences deserve investigation. They do not make every new dashboard or content recommendation an advance. More mentions need not mean more suitable customers. A rise after a website update does not establish that the update caused it.
This is the distinction we are exploring through Signal: not simply whether a business appears, but whether AI understands when it is genuinely relevant and brings it into the conversation for the right reasons.
Update, 29 September 2026: AIDO Labs is now Signal.
For now, the practical foundations remain close to SEO, content marketing and PR. The stronger claim—that optimisation can improve how accurately an assistant matches a business to someone’s actual need—requires separate evidence. The question is whether this work improves that judgement, or mainly gives the publicity industry another way to influence it.