AI answers keep recommending businesses that closed, got bought, or changed their name. Here is why it happens, how much traffic it eats, and how to take that spot.
The short version
Blockbuster peaked in 2004 with 9,094 stores and 84,300 employees. There is one left, in Bend, Oregon, running a DOS-based IBM system. Twenty years after the collapse, “video store” still pulls up “Blockbuster” in a normal human brain. AI answer engines have the same reflex, except they apply it to your industry, your city, and your competitors.
Firms that closed in 2019 still get named as options. Agencies that were acquired and folded into a parent brand still get recommended under the old name. Practices that merged, rebranded, or died quietly still take up slots in a ten-name answer. Meanwhile a living business with a phone that rings does not get mentioned at all.
The reason is simple and a little bleak: there is no obituary column for companies. When a business fails, almost nobody publishes the sentence that says so. The old reviews stay up. The old directory listings stay up. The old “top 10 in your city” roundup from 2021 stays up and keeps getting scraped. The last thing the internet said about that company was that it was good at its job, so that is the fact that survives.
Why the machines get it wrong
Two separate failures stack on top of each other.
1. Training snapshots freeze time. A model learns from text collected up to a cutoff date. It does not receive a notification when a company shuts down. What it has is a last-known-good picture, and that picture never gets corrected unless the open web corrects it first.
2. Live retrieval inherits stale pages. Even engines that search in real time are reading the same uncorrected pages. Columbiaโs Tow Center ran 200 tests across eight AI search tools and found incorrect citations in over 60% of them. Expert reviewers in a separate study flagged misattribution, misrepresented sources, and missing citations as routine, not rare.
Freshness decides who wins the slot. One 2026 analysis found 62% of the most-cited pages in AI answers were published in the previous six months. Semrush found pages not updated quarterly are about 3x more likely to lose their AI citations. If no recent page says a company is gone, the older page saying it is great keeps the seat.
The scale: businesses disappear constantly and quietly
1. BLS data shows 22.1% of new U.S. businesses close within their first year. In the latest cohort that was 218,861 closures, roughly 600 per day.
2. By year five, 48.6% are gone. By year ten, 65.3%. Roughly two out of three names on any list from a decade ago are no longer answering.
3. Acquisitions erase brand names even faster than failure does. PitchBook counted 50,810 M&A transactions globally in 2025, worth close to $5 trillion, with deal count up 12.4% year over year.
4. S&P Global counted 51 deals above $10 billion in 2025 alone, totaling $1.1 trillion. Every one of those swallowed brand names that customers still search for by the old name.
Nobody sends a correction notice. The company just stops answering email, the website expires or redirects, and the reputation keeps floating around unattached to anything real.
Why this costs more than it used to
Two years ago a stale listing was a nuisance. Now it is a channel.
1. AI Overviews appeared on roughly 18% of Google searches in early 2025, about 25.8% by late 2025, and studies in 2026 put the figure anywhere from 48% to 60% depending on method and query mix.
2. Around 68% of U.S. Google searches ended without a click in early 2026. When an AI Overview shows up, the no-click rate runs 80% to 83%.
3. Ahrefs measured a 58% drop in clickthrough for the number one organic result when an AI Overview is present. Seer Interactive put the organic CTR decline at 61%, and found cited brands earn about 35% more clicks.
4. ChatGPT reported 900 million weekly active users in February 2026, up from about 400 million a year earlier.
5. Being cited is not the same as ranking. OrganikPI research found 76.95% of URLs cited by AI engines were not in Googleโs organic top 10.
Put those together and the answer box is the destination for a large and growing share of questions. If a company that no longer exists is inside that box and you are not, you lost the query before anybody clicked anything.
The field report: how to count the ghosts in your category
This takes an afternoon and it will tell you more than most paid audits.
1. Pick one category and one geography. “Commercial roofers in Pittsburgh.” “Family law firms in Tampa.” Something a real buyer would actually type or say.
2. Ask five engines the same three questions. ChatGPT, Gemini, Perplexity, Copilot, and Google AI Mode. Ask for the best providers, the biggest providers, and who to call for a specific problem.
3. Write down every company name that appears. Do not filter yet. Note which engine named it and in what position.
4. Check each name against reality. Secretary of State business registry, current phone number, active website, Google Business Profile status, recent reviews. Five minutes each.
5. Tag each one: alive, closed, acquired, or rebranded. Acquired and rebranded count as ghosts too, because the name being recommended no longer routes anyone anywhere useful.
6. Count the ratio. If three of ten names in a ten-name answer are ghosts, 30% of the available recommendation space in your category is being held by companies that cannot take the call.
What you publish to fix it
This is the uncomfortable turn, so here it is plainly: your dead competitor is a distribution channel. Somebody has to publish the sentence that says where those customers went, and there is no reason it should not be you. Handle it like a professional, not a vulture. State facts, do not gloat, and do not invent anything.
1. Write one dated, factual page per ghost. “Harmon & Reese closed in March 2023. Their commercial clients in Allegheny County have largely moved to three remaining firms, including ours.” Name the date. Name the change. Keep it short.
2. Answer the question the searcher is actually asking. Where do I get my records? Who honors the old warranty? Is my deposit gone? Who does this work now? Those pages get cited because they resolve the query.
3. Put freshness signals on the page. A visible Published and Last Updated date, a Last-Modified or ETag header, a stable changelog URL, and explicit version or effective dates on anything that changes.
4. Do not delete the old stuff. If you acquired the firm, keep the legacy pages alive with a clear “Deprecated” or “Now part of” notice and a redirect. Deleting a page does not remove the claims about it living on other sites.
5. Get it said somewhere other than your own site. Muck Rack found 84% of AI citations come from earned media rather than brand-owned pages. A trade publication, a local business journal, a Wikipedia or Wikidata correction, and an updated registry entry are worth more than another blog post.
6. Recheck quarterly. Retrieval-heavy engines pick up changes within days. Training-baselined engines need the next cycle. Both need the update to exist somewhere first.
What not to do
1. Do not claim you acquired a firm you did not acquire. That is fraud with a marketing budget attached.
2. Do not write “X went out of business” without confirming it. Slow-moving small firms look dead online all the time and are not.
3. Do not build a wall of thin pages about every competitor. Five real pages beat fifty generated ones, and the generated ones will get you filtered.
4. Do not stuff keywords. Research found citing sources, adding quotes, and including statistics each lifted AI visibility by 30% to 40%, while keyword stuffing lowered it and schema markup did roughly nothing.
The point
Blockbuster earned its ghost. It was a 9,000-store cultural institution that turned down Netflix for $50 million in 2000, and the world decided to keep telling the story. The three defunct competitors sitting in your categoryโs AI answers earned nothing. They are there because the internet never got around to saying otherwise, and because you did not either.
Go count them. Then write the sentence.
Sources: BLS Business Employment Dynamics via LendingTree (2026); PitchBook 2025 Annual Global M&A Report; S&P Global Market Intelligence Q4 2025; Tow Center for Digital Journalism, Columbia (Jaลบwiลska & Chandrasekar, March 2025); Ahrefs; Seer Interactive; SparkToro/Similarweb; Semrush; Muck Rack (May 2026); OrganikPI; Conductor; OpenAI; Roadtrippers and NBC News on the Bend store.

Discover more from 1000 Startups
Subscribe to get the latest posts sent to your email.
Claude Penland builds the marketing and business strategy for companies that are good at what they do and hard to find. Thirty years operating, one exit, eight of them as a practicing casualty actuary.
The free two-page read is genuinely free. Email claude@1000startups.com and I'll send back what I can see from the outside. Or see the work samples and how to work with me.