Search your own service in Google and you may well sit at position one. Ask ChatGPT the same question, the way a buyer would phrase it, and three other businesses get named instead. Nothing has broken. You are ranked but not recommended, and those are now two separate things. In the first four months of 2026, 68.01% of US Google searches ended without a click, according to SparkToro’s analysis of Similarweb clickstream data published in June 2026. Not every one of those was lost business. Enough of them were. Here is what the gap is, why it stays quiet, and how to check yours in about a minute.
Key takeaways
- Ranking and being recommended are different outcomes with different inputs. You can hold one and lose the other.
- The loss is quiet. Positions hold, traffic looks roughly normal, and enquiries thin out slowly.
- No AI platform publishes how it selects the businesses it names, so this is observed behaviour rather than confirmed mechanics.
- Your buyers have not changed. They still want what you sell. Only the place they shortlist has moved.
- This is checkable in minutes, on your own business, without buying anything.
- Nobody can guarantee an AI recommendation. What can be closed is the reason you are being left out.
What does ranked but not recommended actually mean?
Ranked means a search engine has placed your page in a list and a person can choose to click it. Recommended means an AI assistant has named your business inside an answer, usually alongside two or three others, before the person has visited anyone’s website. The first is a position. The second is a shortlist place.
The two are earned differently. A ranking rewards relevance, authority and a healthy technical base. A recommendation appears to reward something narrower: whether a model can identify your business confidently, describe it consistently, and find other people saying so.
That difference is why a strong SEO position offers no protection here. It was never built to.
| Ranked | Recommended | |
|---|---|---|
| What it is | A position in a list of links | A named mention inside a generated answer |
| Where the buyer sees you | On a results page they scroll | In the answer, before any website |
| What the buyer does next | Chooses between ten options | Chooses between three or four |
| What appears to decide it | Relevance, authority, links, technical health | Entity clarity, consistency, third-party mentions, reviews |
| How you measure it | Rankings, sessions, conversions | Mention and citation rate across assistants |
| What it is worth | Traffic you can then convert | A place on the shortlist before the traffic exists |
Why do rankings hold while enquiries quietly thin out?
Because the traffic you lose first is the traffic you never see arriving. A buyer who gets a usable answer on the results page, or inside an assistant, never becomes a session in your analytics. Your rankings report looks fine, because the ranking genuinely is fine.
Pew Research Center measured the click side of this in March 2025, across 68,879 Google searches from 900 US adults, and published it in July 2025. Where an AI summary appeared, people clicked a traditional result in 8% of visits. Where no summary appeared, they clicked in 15%. Links inside the summary itself were clicked in 1% of visits.
Both of those studies are US data. No equivalent UK measurement has been published, so treat the direction as reliable and the exact number as American.
The commercial version is simpler. The same position now produces fewer enquiries than it did, and nothing in your reporting tells you why.
What is an AI assistant doing when it names three businesses?
It is assembling an answer from what it can read and verify, then naming the businesses it is most confident about. It is not consulting a ranking table. There is no position one to occupy inside a generated answer.
From what can be observed, the businesses that get named tend to be easy to identify, described the same way in several places, mentioned by sources other than themselves, and visible in reviews. The ones left out are usually not worse businesses. They are harder to pin down.
This is worth stating plainly, because the subject attracts confident claims. No major platform publishes its selection mechanism. Anyone telling you exactly how ChatGPT picks is describing a pattern, not a documented rule, and so are we.
The practical consequence holds either way. If a model cannot confidently say who you are and what you do, it will name a business it can.
Does this mean your buyers have changed?
No. This is the part most owners get wrong, and it is the reassuring part. A facilities manager still needs a cleaning contractor for three sites. A landlord still needs a managing agent. The demand is unchanged, the budget is unchanged, and the reasons they would choose you are unchanged.
What moved is the shortlist. It used to be built by scrolling. Now a growing share of it arrives pre-built, with three names on it.
That reframes the work from defending traffic to earning a place in the answer. It is the same commercial goal as SEO has always had, which is why this sits as the next stage of search rather than a replacement for it. If you want the distinction between the acronyms set out properly, we have covered that separately.
What did Blockbuster actually get wrong?
Not the technology. Blockbuster understood mail and streaming perfectly well. What it could not do was act against the part of the business that was still paying the bills, because every incentive inside the company pointed at protecting the stores that were still busy.
The lesson is about incentives and discovery, not stupidity. When your existing channel is still performing on paper, there is no obvious moment to look somewhere else, and the signal that something has shifted is always weaker than the signal that things are fine.
Ranked but not recommended works the same way. Your Google position is still paying you something. That is exactly what makes the gap easy to ignore.
How do you check your own business in under a minute?
Open an AI assistant and ask it what a buyer would ask, then record what comes back. You are not looking for a score. You are looking for whether your name appears at all.
- Write down the three questions a buyer asks before they know your name. Not your brand, your service and your area. For example, the best office cleaning company in Croydon.
- Ask all three in ChatGPT, then repeat them in Gemini and Perplexity. The answers differ, and that difference matters.
- Note who gets named. Write down the competitors, not just whether you appeared.
- Ask a follow-up as a buyer would: who is the most reliable of those, and why. Watch which evidence the answer leans on.
- Do it again in a fortnight. One answer is a snapshot. Two are a pattern.
If you are named, you have a position worth protecting. If you are not, you now have a list of the businesses being recommended instead of you, which is more useful than any ranking report. Our free AI visibility check runs a structured version of this and returns a score, the prompts tested and what is blocking you, in about thirty seconds.
What do you do with what the check tells you?
Fix the reasons you are being left out, in the order that costs least. Usually that means making your key pages answer questions directly, making your business details identical everywhere they appear, and building mentions on sites you do not own.
None of that is exotic, and most of it improves your ordinary search performance at the same time. That is the honest version: this is not a separate budget so much as a change in what the existing budget should be pointed at.
What it is not is a guarantee. Nobody can promise you a citation in ChatGPT or a place in an AI Overview, and anyone who does is making the claim we built this agency in reaction to. What can be promised is measurement, so you can see whether mentions move.
The full argument, the sector examples and the campaign this article belongs to sit on our Ranked Not Recommended page.
Who should be doing this work?
Somebody senior enough to tell you when the answer is to spend less. This is a new field with a lot of confident noise in it, and the useful judgement is knowing which of the five possible fixes actually applies to your site.
Our founder spent seven years in-house across facilities management and property before this agency existed, and over ten years in B2B marketing. He has sat on the buying side of exactly this decision, holding an invoice for search work that was not producing enquiries. The first time he ran this check on a business he knew ranked well for its main commercial terms, it was not named once.
The results behind that experience include organic visits moving from 40 a month to well over 1,000 on commercial terms inside one month, and an organic channel built from nothing to over £1M a year in a commercial cleaning business. These results were achieved by Marcus Lobow, some in-house as a senior marketer and some through Lobo Digital.
At Lobo Digital the person you brief is the person who does the work. Our pricing is published at £62.50 per hour plus VAT, at 15, 25, 40 or 65 hours a month with a three month minimum, which is £937.50 a month at the lowest band. Our AI search optimisation service and the deeper AI visibility audit both start from what the check finds, not from a fixed task list.
Frequently asked questions
Does this mean SEO is dead?
No. AI assistants and AI Overviews draw heavily on crawlable, well-structured, credible pages, which is what good SEO produces. What has changed is the outcome you measure it against. A ranking that produces no enquiries is no longer evidence that the work is paying. Being cited is becoming as valuable as being ranked, and the two share most of the same foundations.
How is being recommended different from ranking?
A ranking places your page in a list and waits for a click. A recommendation names your business inside an answer, usually alongside two or three others, before anyone reaches a website. Rankings reward relevance and authority. Recommendations appear to reward how clearly and consistently your business can be identified and verified by sources other than you.
Does this affect small businesses or only large ones?
It affects small businesses more, in both directions. An AI answer names three or four businesses rather than listing ten, so a smaller firm with clear content and consistent information can appear next to far larger competitors. The same compression means being absent costs more, because there is no second page to be found on.
How would I know if it is happening to me?
Ask an assistant the questions your buyers ask, in their words rather than your brand name, and see who gets named. Do it across ChatGPT, Gemini and Perplexity, because they answer differently. If competitors appear and you do not, that is the gap. Repeat it a fortnight later to tell a pattern from a one-off answer.
Can anyone guarantee my business gets recommended by AI?
No, and we will not. AI answers are generated rather than ranked, so there is no position to buy or hold. What can be done is removing the reasons you are excluded, which is unclear content, inconsistent business information, thin third-party mentions and pages that AI crawlers cannot read, then reporting monthly on whether mentions move.
Run the free AI visibility check
Find out whether ChatGPT, Gemini and Perplexity name your business when your buyers ask, or name someone else. It takes about thirty seconds, there is no signup to see your score, and you keep the findings either way.
Run the free AI visibility check
Sources: SparkToro, June 2026; Pew Research Center, July 2025.
