The free citation is losing room on the screen. In three years, AI search went from answering questions with cited sources to putting sponsored placements beside, inside and below those answers.

Put the dates in order and the pattern is hard to miss: an answer surface opens with citations, then makes space for ads. That does not mean an advertiser can buy its way into a source list. It means a business can earn a citation while the surrounding layout grows more crowded. The old search results page followed a similar path, with paid placements claiming more of the space a buyer sees first. This timeline tracks where that happened in AI search, what changed for practitioners and why a citation checklist needs a date on it.

December 2022 to February 2023: Perplexity makes the citation part of the answer

Perplexity launched its flagship search engine on December 7, 2022 as a conversational engine that synthesized web results and returned answers with citations attached. The consequential choice was to print sources with the answer rather than send the user straight to a list of links.

The mechanism is simple enough: retrieve pages, compress the useful material into an answer, then show the sources. For a business, that creates a second way to be found. A page does not have to be the first blue link to supply a sentence the buyer reads in the answer. It does, however, have to give the engine something clear enough to use. A page that circles the question for six paragraphs before stating its point makes that job harder.

The early record here is about an answer format, not proof that citations produced a particular volume of traffic. That distinction matters. Ranking and being referenced became different jobs, even before anyone could say how much the second job was worth. From this point on, I would check both: where a page ranked and whether the answer named it.

Microsoft began rolling out the new Bing with Bing Chat on February 7, 2023. The interesting part for a search operator was not the chatbot label. It was the footnotes attached to its answers.

The engine retrieved pages, wrote a response and displayed sources the reader could follow. That changed the order in which a buyer encountered the information. Instead of scanning rankings first and deciding which page deserved a click, the buyer could read an answer first and click a citation for detail. A strong position in the classic results still mattered, but it no longer described the whole opportunity.

I used to treat a good snippet at position three as a useful consolation prize. In an answer-led layout, the more useful question is whether the page says something the engine can carry into its response. The citation is a separate piece of visible real estate. Check the answer itself before declaring a ranking win.

Google announced Search Generative Experience at Google I/O on May 10, 2023 as an opt-in experiment in Search Labs. Its AI snapshot appeared above the classic links, summarized information and included links within the snapshot. This was a test, not a universal replacement for search results. It still showed practitioners what could move to the top of the page.

Ranking helped because SGE drew on search results, but ranking alone did not settle which material appeared in the summary. A page that answered a question in an extractable passage gave the system more to work with than a page padded with generic claims. The practical change was not a new incantation called “AI SEO.” It was a sharper reason to write the answer plainly, near the point where the reader needed it.

The page now had to work for two readers: the person who clicked through and the system assembling an answer before that click. If your best explanation was buried below the sales pitch, move the explanation up.

May 14, 2024: AI Overviews put the answer in the main results

Google began rolling out AI Overviews to everyone in the U.S. during the week of May 15, 2024, using a custom Gemini model tied into its Search systems. What had been an opt-in experiment became part of the main search experience.

An Overview put a generated response at the top, with links beside it. For an informational query, the buyer could get a summary without scrolling through the familiar list. A citation inside that summary therefore occupied a different position from an ordinary page-one listing. The listing had not vanished. It was simply below the thing the buyer read first.

Early wrong answers embarrassed Google, and it pulled back on some query types. I read that as a reason to inspect actual queries, not to assume the feature was going away. An Overview appearing for one buyer question tells you more about that question’s layout than any general claim about AI search does. Ranking without a citation can leave you beneath the answer. Check the results for the questions that bring you business, not just the rankings in a report.

May 21, 2024: Google announces ads for AI Overviews

At Google Marketing Live on May 21, 2024, Google announced that Search and Shopping ads would soon appear inside AI Overviews in a labeled sponsored section when relevant to the query. “Would soon” matters: this was an announcement of a paid placement, not evidence that every Overview already carried one.

The commercial logic was plain. Once an Overview occupied the top of the results page, a sponsored section within it could occupy space beside the generated answer. Organic citations did not become purchasable. The layout around them became sellable. Anyone who has watched sponsored results take priority on a search page recognizes the direction, even if the format looks different.

That split creates two jobs for a practitioner. Make the page useful enough to earn a citation, and decide separately whether the relevant query justifies an ad. One does not substitute for the other. The free citation had a planned paid neighbor within a week of the wider rollout. Keep the citation plan and the media budget separate, even when both appear on the same screen.

October 31, 2024: ChatGPT Search opens another citation surface

ChatGPT Search launched October 31, 2024 inside ChatGPT for Plus and Team users, with a search icon and inline citations linking to sources. The familiar mechanism appeared in a different setting: retrieve current material, compose a response and show where it came from.

The question itself could look different there. A buyer might ask for the best payroll tool for a 20-person agency that bills hourly rather than type a short category keyword. That longer prompt gives a useful page several chances to be relevant: the buyer’s size, work and constraint can all be answered in plain language. It also exposes thin copy quickly. Repeating “best payroll solution” does little for someone who has described the actual decision.

This was a new place to check citations, not a reason to assume Google and ChatGPT would select the same sources. Citation work split by engine. If buyers ask both surfaces about your category, inspect both answers and write for the full questions they ask, not just the short keywords you already track.

November 12, 2024: Perplexity puts sponsored questions beside answers

Perplexity began experimenting with ads in the U.S. in November 2024. Its sponsored follow-up questions sat beside answers and carried a sponsored label. Launch partners included Indeed, Whole Foods, Universal McCann and PMG.

The design is worth looking at closely. A sponsor could place a question near the answer, but the answer that followed was still generated by Perplexity’s AI rather than written by the sponsor. Advertisers did not get access to personal information. Cause first: the source-backed answer remained the engine’s work. Effect: a brand could pay for position near it without buying the wording of the answer or the citation.

That is the distinction I would put at the top of any budget conversation. You could not pay to be cited; you could pay to sit beside a citation. Pages had to earn their place in the answer. Ads needed their own case, based on whether the buyer question was worth paying to appear near. Calling both activities “AI visibility” may look tidy on a deck. It does not tell anyone which work to do.

Paper craft timeline showing citation space alongside growing paid ad blocks

Early 2025: AI Mode makes the long buyer question harder to ignore

Google pushed its conversational AI Mode into wider testing through spring 2025. In that format, a buyer could put budget, location, constraints and timing into one question instead of reducing the problem to two keywords. The answer engine could hold that context, so the page had to address more of the request to be useful.

Say you sell garage doors in Dallas. The short query is garage door repair. The longer one is who can replace a double garage door in Dallas this week without upselling me a full opener. A page with prices, service area and a straight account of the work has material the answer can use. A page that says only that you provide “quality solutions” has given it almost nothing. I would rather edit that page than spend an afternoon inventing a new keyword label for the same problem.

This does not make every long prompt worth a new page. It changes what to look for in the questions your buyers already ask. Match the whole question, not just its category keyword. Put the answer, the relevant constraints and the next step where a reader can find them without digging.

November 2025 to February 2026: ads reach deeper into the response

Ads in Google AI Mode were spotted on November 20–21, 2025 for local and service queries. They were labeled sponsored and appeared at the bottom of the response. Google confirmed a test but gave no full rollout date. The distinction matters: a spotted test shows where an ad can fit, not how often every buyer will see one.

OpenAI then started testing ads in ChatGPT in February 2026 for logged-in Free and Go users in the U.S. Sponsored units sat below and separate from the answer; sensitive topics and under-18s were excluded. Again, the paid block did not rewrite the response. It occupied space in the same experience where the buyer was reading it.

The two tests did not make the products identical. They did extend the same commercial pattern to more answer surfaces. A citation can remain unpaid while the screen around it becomes paid. For a practitioner, that means checking the placement itself: where is the citation, where is the sponsored unit, and what does the buyer see before either one?

October 2026: treat an earned citation as a position worth defending

Here is my reading of the line from the first cited answers to the latest ad tests. AI search has not replaced organic citations with ads. It has kept the citations and made room to sell beside them. That is close enough to the old search-results pattern to take seriously, without pretending every engine has the same layout or rollout schedule.

It is also why ranking alone no longer wins the citation. A ranked page still has value, but the buyer may encounter a generated answer before the list. The history of paid placements moving into AI Mode points to the same practical problem: earning a mention and buying nearby visibility are separate decisions, and the space for both keeps changing.

The businesses earning organic citations now are securing a position before the paid layer grows around it. That does not mean every business needs a citation campaign. If your buyers do not research before contacting you, skip this work and focus on the ad that reaches them. If they do research, I would take three steps:

  1. Find the buyer questions. Pick 10 to 20 of the longer prompts your best customers ask, including budget or location where those details matter. Check who gets cited in ChatGPT, Perplexity and AI Overviews. Write a page for each question worth answering, and answer it in the first 80 words.
  2. Make the answer extractable. Give each passage one clear job. Name the product and place plainly; state prices and steps outright where they belong. If a machine cannot lift a clean sentence from the page, rewriting the headline will not fix the missing answer.
  3. Track each engine separately. Google’s wider AI Overview rollout came in May 2024; ChatGPT Search launched in October 2024; AI Mode ad tests appeared in late 2025. Check citations weekly by engine. An average hides the surface that changed while you were watching the others.

Old checklists show how quickly AI Overview citation advice can expire. A guide without a date may describe a different layout from the one your buyer sees this week. My rule is plain: check the date, then check the live answer for your money prompts. The citation may still be free. Its surroundings are not guaranteed to stay that way.

That is why I want paid and organic work connected at groas. The team manages Google Ads and ChatGPT Ads around the clock alongside citation work: pages written to be lifted, citations tracked by engine and changes logged with a reason. Say you spend $20k a month and a buyer sees your rival cited with your rival’s ad below the answer. You have two problems on one screen. One team should own both. Earn the citation while you can, because the timeline says you may soon be bidding for more of the space around it.