Google Can Rank Without an Index. What Should Contractors Do?

Not yet, and not in the way the headlines suggest. Google researchers proved on paper that one AI model can rank documents without a separate index, but nobody has swapped it into real search. The lesson for contractors: stop obsessing over position three and start earning a place in the answer.

What did Google actually prove about ranking without an index?

In a January 2026 paper, Google researchers proved that a single language model with a fixed size can rank an unlimited number of documents without a separate index. According to Duane Forrester at Search Engine Journal, it is a theoretical result, not a shipped product.

Here is the plain version. Today, search works in two stages. A fast, cheap system pulls a big pile of candidate pages. A slower, smarter system then sorts that pile. The DeepMind team proposes folding both jobs into one model that simply produces the ranking as it goes.

Forrester is honest about how small the experiments were. The team fine-tuned Mistral-7B, which is not a frontier model. They tested on a shopping set of 310 examples and on WordNet, a database of about 82,000 English nouns. Forrester calls it "a foundation, not a deployment."

I like that he says so. A lot of people will read this paper and announce that SEO is dead. It is not. But the direction matters, and the direction is where a contractor should pay attention.

Are the ten blue links actually going away?

Nobody outside Google knows. Forrester's argument is that the ranked list stops being something you see and becomes something the model works out privately on the way to an answer. That is informed speculation, and he says so. The research history behind it is real, though.

Forrester lays out the arc. In 2021, Google researchers published a paper arguing a search system should answer directly from a corpus it can cite instead of handing over a list of references. In 2022, a follow-up showed a single model could map a query straight to document identifiers. The 2026 paper supplies the math.

Proposal, prototype, proof. Five years of research pointing the same way is not an accident.

My take: I would not bet your business on the list vanishing next year. I also would not bet on it looking the same in five. If you run a roofing or plumbing company, the cheap move is to prepare for both.

What does being included in the answer mean for a home-services business?

It means that when a homeowner asks an AI assistant or a search box for help, your business is one of the few names that comes back. There may be no page of ten results to scroll. Being included beats being ranked, because a list can show ten names and an answer may show one.

Think about how a homeowner actually behaves. A pipe bursts. They are not browsing ten results and comparing. They want a name and a phone number, fast. If an AI hands over two or three businesses, the contractor sitting at position six on a list nobody sees is invisible.

That is the shift. Ranking is a position. Inclusion is a yes or no.

I will be careful here. Nobody has shown that Google will present local results this way, and I am not going to pretend otherwise. But AI assistants already answer local questions with a short list of names. That is where homeowners are heading, so that is where I would aim.

How do I make my business easy for an AI to pick?

Make your business clearly distinct, clearly described and clearly current, everywhere it appears online. The paper suggests a model has to tell your business apart from its neighbors. If you look identical to the competitor down the road, you give it no reason to pick you.

Forrester makes a point I find useful. In this kind of system, a document is not a URL. It is a short code the model generates, and similar items share similar codes. If a page cannot be separated from its neighbors, it cannot hold a distinct identifier. Two near-duplicates compete for the same address, and one of them loses outright.

Translate that to a contractor. Think of the website with forty city pages that are identical except for the city name. That is a near-duplicate problem. That one is my inference, not something the paper tests, but it fits.

Here is what I would do, in order:

1. Say exactly what you do and where. Use the same plain wording for your services and service areas on your Google Business Profile, your website and your directory listings. "Emergency water heater replacement in Plano" beats "quality home solutions."
2. Give every page one job. One page for roof repair, one for roof replacement, one for storm damage. Do not clone a page forty times with a new city name.
3. Keep your profile current. Hours, services, photos, holiday closures and posts. Forrester notes that new information has to be learned in a model like this, and that Google's own follow-up work found models forget older material when learning new material. Stale details are a risk either way.
4. Earn reviews that describe real jobs. A review that says "replaced our furnace on a Sunday in January" tells a machine far more than "great service!"
5. Check what the machines see. Start with the free Local Visibility Score to see where you stand, then run the AI Visibility Report to see whether AI tools mention you at all.

None of this is exotic. It is the same discipline that has always worked for local search, pointed at a future where the machine decides instead of the searcher.

Does the ranking function matter if the model computes the list itself?

Yes, because it still has to learn what a good result looks like from somewhere. Forrester points out that the training data becomes the ranking algorithm. Whatever produced the graded lists, whether click logs, rater judgments or engagement signals, is what the model reproduces.

Read that again as a contractor. Click logs and engagement signals are what real people do. They call, tap directions, read reviews, ask questions, book. In that kind of system, the businesses real customers keep choosing are the ones the model learns to prefer.

That is why I keep telling owners not to treat their Google Business Profile as a form to fill out once. It is the place where customer behavior and your business details meet. A profile with current services, fresh photos, answered questions and recent reviews gives real customers something to act on. That behavior is the raw material.

You cannot reverse engineer a hidden list of ranking factors. Forrester says there is no separate layer to chase. What you can do is be the business people keep choosing. If you want a second pair of eyes on yours, the Google Profile Audit is a good place to start.

Should I stop caring about my Google ranking?

No. Rankings still drive calls today, and this research is a foundation, not a launch. Keep tracking where you show up for your main searches. Just stop treating a ranking as the finish line, because the goal is shifting from a position on a list to being named at all.

There is one more thing worth saying. Forrester disclosed that he founded CitationIQ, a platform that measures AI visibility, so he has a stake in the claim that measurement is changing. I appreciate the disclosure, and I will give you mine: LeadsPass manages Google Business Profiles and local search, so I also have a stake in telling you profiles matter.

Judge the advice on its own merits. It costs little to describe your business clearly, keep it current and earn honest reviews. If Google never changes a thing, you still win. If it changes everything, you are ahead.

Common questions

Has Google replaced its search index with an AI model?

No. The January 2026 paper proves a theoretical capacity and shows a training method on two small datasets. According to Duane Forrester, nobody has replaced a web index with it. It signals where Google research is heading, not what is running behind your searches today.

Will the ten blue links disappear?

Nobody knows. Forrester argues the ranked list becomes something the model computes internally on its way to an answer, but he calls this informed speculation. Plan for both outcomes: stay visible in rankings now, and build the clarity and reputation that get you named in answers.

Is my Google Business Profile still worth the time?

Yes. It is where your services, location, hours, photos and reviews sit in one place, and customer behavior around it is exactly the kind of signal a learned ranking system would draw on. A complete, current profile helps under today's system and plausibly under the next one.

How can I tell whether AI tools mention my business?

Ask the tools the questions your customers ask, such as who is the best emergency plumber in your town, and note who comes back. For a faster read, the free AI Visibility Report checks how your business shows up and where the gaps are.