This part of the shift confuses people more than any other.
A client came to us this year with a reasonable complaint. Their SEO was in good shape: they ranked on page one for their main service terms in their market, and their traffic was steady. Yet when they asked ChatGPT who to hire for the exact service they provide in the exact city they operate in, they were not mentioned once across a dozen attempts.
They assumed something was broken. Nothing was. The two systems do different jobs, and their marketing had only ever been built for one of them.
Google ranks pages, AI cites companies
A search ranking is a judgment about this page relative to this query. Is this page a good response to these words? Relevance, authority signals, structure, speed, and roughly two hundred other factors get applied to one document at a time.
An AI recommendation is a judgment about this business relative to this need. The model is reasoning about a company as a whole, and it builds its picture of that company from everything it has absorbed across the web, weighted by how much it trusts each source.
So your page ranks and your company gets cited. Those are different units of measurement, and optimizing one does not automatically produce the other.
AI discounts what you say about yourself
The second gap is more uncomfortable, and it trips up businesses with beautiful websites. These systems heavily discount what a company says about itself. The discount comes from the training data, where the pattern is obvious: every agency website claims to be results-driven, client-focused, and strategic. A claim that appears identically on fifty thousand sites carries no information, so the model learns to weight it near zero.
Corroboration carries the weight. A trade publication describing what you did. A directory listing that independently states your specialty and size. A client quoted somewhere you do not control. An association membership. A roundup that put you on a list alongside your competitors.
Your website is where you make the claim, and everywhere else is where it gets verified. You can win SEO largely on your own property. Citation mostly happens on other people’s.
Inconsistent descriptions lower your odds
This gap is the easiest to fix. If a model finds your business described several different ways across the sources it trusts, it gets less confident about naming you. One directory says you do web design. Another says public relations. Your homepage says integrated marketing and communications. A conference bio says brand strategy. A person reading those would still see one firm. A system assembling a reliable picture sees noise, and noise lowers the odds of a confident recommendation.
We found exactly this when we audited ourselves. TSN appeared across a scatter of old listings with a scatter of old descriptions, several of them years out of date and a couple naming services we no longer lead with. The model had no stable picture of us to draw on, so it drew on someone else.
How to fix your listings
The work is unglamorous, which is why most businesses skip it.
Start by finding every place your business is described that you do not control: directories, association listings, chamber profiles, old press mentions, partner pages, review platforms. Most businesses are startled by how many there are and how stale they have become.
Then write one canonical description of what you do, who you do it for, and what makes you specifically different. Keep it to two or three sentences, concrete enough that a stranger could repeat them.
Then make every one of those listings say exactly that, word for word.
It is tedious work, and it moves the needle more than another blog post will.
How long this takes to pay off
This will not pay off next week. These systems absorb the web on their own schedule, and a corrected directory listing may take months to show up in how a model describes you.
That lag is the reason to start before you can prove the return. The businesses that will be visible eighteen months from now are putting the corroboration in place now, while their competitors wait for someone to show them a case study.
Next week we look at which of those third-party sources actually carry weight. They are not equal, and some of the ones agencies chase hardest matter least.
Frequently Asked Questions
How can a page rank well on Google but never be mentioned by AI?
Because the two systems measure different units. A search ranking judges a single page against a single query. An AI recommendation judges a business as an entity, assembling its picture from sources across the whole web. Winning the first does not automatically produce the second.
Why is our own website discounted?
Self-descriptive marketing claims appear near-identically across tens of thousands of company sites, so they carry little distinguishing information. Systems learn to weight them lightly and to rely instead on third-party corroboration such as trade press, directories and independently published client commentary.
What is the consistency problem?
When a business is described differently across the sources a model trusts, for example web design in one directory and public relations in another, the model has no stable picture to draw on. That inconsistency reduces the likelihood of a confident recommendation.
What is the first practical step to fix this?
Inventory every place the business is described that you do not control, write one canonical two or three sentence description of what you do and for whom, then update every listing to match it exactly rather than approximately.
How long does it take to see results?
Months rather than weeks. These systems absorb the web on their own schedule, so corrected listings surface gradually. That lag is the argument for starting before the return can be proven, not against it.