A short history of AI in marketing: lessons for small businesses

Before choosing AI, distinguish the job it does

A customer list, a scheduled email and an AI-generated article all use software, but they solve different problems. Keeping these distinctions clear helps you understand a proposal. Does the system store information, follow a rule, predict an outcome or generate a draft?

For example, a fixed reminder sent after a form submission is automation. Predicting which ad auction may lead to a sale is a different capability. Neither tells you whether the offer on the landing page is persuasive.

This is a selective history of developments relevant to marketing, not a claim that every earlier statistical or digital tool was AI.

For a practical next step: How to measure AI search visibility—and connect it to leads and sales.

2016: conversion-based bidding gets a clearer identity

In July 2016, Google introduced the Smart Bidding name for conversion-based automated bidding. The announcement described machine learning and bids adjusted for individual auctions.

The lesson for an owner: ask which outcome the advertising system is being asked to optimise. A form submission, a qualified enquiry and a completed sale represent different stages. Your campaign brief should explain which actions are measured and how their quality will be reviewed.

Automatic bidding still leaves decisions for the business: the offer, budget, product economics and what counts as useful progress.

2019: search improves its handling of context

In October 2019, Google announced the application of BERT models to Search, describing improvements in how it interpreted the context of words in a query.

The lesson for an owner: bring actual customer questions into your website brief. A product page should explain who the product suits, relevant differences and what buyers need to decide. A service page should answer the questions people ask before requesting a quote.

That is a practical editorial direction, not a formula for a guaranteed ranking.

2023–2024: generated answers become part of search

Google announced generative Search experiments in May 2023. Its August 2024 update recorded the launch of AI Overviews that May. These are documented product milestones, not evidence that all buyers changed their behaviour at the same time.

The lesson for an owner: observe how your own category is presented. Save representative buyer questions and note whether answers cite relevant sources. Do not assume an AI mention produces a visit or a sale.

Generated content also changes production work. A useful agency brief should explain which drafts or variations are needed, who checks them and what makes them ready to publish.

Use the history to ask better questions today

The age of one technology does not establish the reliability of every newer tool. When a proposal uses “AI” as its main selling point, ask the agency to identify the actual work:

• What information do you need from our business?
• What will we receive and be able to review?
• Which decisions remain with us?
• What measurement is available, and what will it not establish?
• Who implements the changes after the analysis?

An illustrative content project might use your product information to draft a comparison page, then check and publish the agreed version. That is a defined piece of work. Describing it as “AI transformation” adds little to the buying decision.

Digital Intelligence offers website, content, SEO and campaign work with AI used where appropriate to the task. The proposal should make the scope clear, including any specialist involvement.