LinkedIn
Google put an answer above the blue links, then put ads in the answer. That sequence should change how you spend on search.
Since the 2023 Search Generative Experience preview, each turning point has pushed attention away from the familiar list of results and toward a smaller set of cited sources. Ads inside AI Overviews show that Google now prices space on that answer surface. An organic citation is the seat you do not pay for on every click. Read the changes in order, and waiting looks less like caution and more like an expensive habit.
May 2023: SGE puts answers above the links
Source cards reveal a different way to be seen
On May 10, 2023, Google introduced the Search Generative Experience (SGE) at I/O as an opt-in Search Labs experiment. Its tinted snapshot synthesized an answer from across the web and took up a large stretch of the top of the results page. Small source cards appeared beside and beneath the text.
The opt-in label made SGE easy to dismiss. A search team could reasonably keep watching its usual rankings and ad positions while the experiment ran. But the source cards mattered more than the toggle: they gave a page a way to appear inside the answer rather than only below it.
A citation was not the same thing as a number-one ranking. The visible result was a new contest for the passages Google could use to answer a query. That did not make rankings worthless, and the Labs preview did not prove what a broad rollout would do to traffic. It did show what the old reporting routine could miss. A page might hold its familiar position while a different source supplied the answer a searcher read first.
For practitioners, that broke a comfortable division of labor. SEO could no longer treat a ranked URL as the complete visibility report; paid search could no longer assume every valuable impression sat in the familiar text-ad stack. The practical question became narrower and more useful: when a buyer asks something specific, does your page give Google a specific answer to cite?
May 2024: AI Overviews make the citation problem mainstream
The odd answers distract from the layout change
On May 14, 2024, Google rebranded SGE as AI Overviews and released it across the United States without the Labs opt-in. The rollout drew ridicule for answers about eating a small rock a day and using non-toxic glue to keep cheese on pizza. Google tightened query filters in response. Those screenshots were memorable. For search operators, the less amusing development was that an AI answer now appeared in ordinary search.
The assumption that page-one dominance guaranteed a place in that answer did not survive scrutiny. In an analysis of 863,000 results pages and 4 million cited URLs, Ahrefs found that roughly 37% to 38% of cited URLs ranked in the organic top 10. Another 26.2% ranked between positions 11 and 100, while 36.7% did not rank in the top 100. Conductor’s analysis of nearly 168 million citations found that pages in organic position one were cited roughly 25% of the time, falling to 10.2% at position 10. More than half its citations bypassed page one.
This changed what a position report could tell you. A top ranking remained valuable, but it was not a receipt for an AI Overview citation. Conversely, a page outside the top results could still enter the answer. If your pricing explanation sat beneath eight paragraphs of category history while another page offered the relevant comparison near the top, the second page was easier to use as a concise source. More backlinks to the introduction would not fix the missing answer.
The response was not to abandon ranking work. It was to inspect the pages already attracting relevant searches and make their useful facts easy to find. Report rankings and citations separately. Otherwise, a stable position can conceal a loss of attention at the top of the page.
October 2024: Google brings ads into AI Overviews
The answer becomes priced space
On October 3, 2024, Google launched ads within AI Overviews for mobile searchers in the United States. Text and shopping units, marked “Sponsored,” appeared with the generated answers. Google drew from eligible automated campaigns, including Performance Max and Search campaigns using Smart Bidding, rather than asking advertisers to run a separate answer-box auction.
The commercial logic was plain. If a searcher can resolve a buying question in the AI Overview without scrolling down to conventional results, the top of that overview is valuable ad space. Google placed paid inventory there. The same answer surface now carried both citations and ads. Organic and paid search might still have separate teams and budgets, but they were competing for attention in the same small part of the page.
Click data made that overlap harder to ignore. Research tracked by Seer Interactive and Pew Research Center found that organic click-through rates on queries with an AI Overview fell roughly 61%, from 1.76% to 0.61%. Searchers clicked traditional organic listings 8% of the time on pages with an AI Overview, compared with 15% on standard results pages. In the cited-versus-uncited comparison reported there, brands cited in the overview earned 35% higher organic CTR and 91% higher paid CTR.
Those figures do not mean every citation produces the same lift or that an ad replaces a citation. They show why a report covering only classic organic clicks, or only paid placement, misses the commercial picture. When the answer appears, fewer searchers reach the old listings. Being named in the answer matters more than holding a position beneath it.
For a PPC operator, the immediate task is to compare the two kinds of presence on the same buyer queries. If your site is cited, know whether your campaigns also show. If it is not, know whether paid placement is carrying visibility you assumed organic had already earned. That is an allocation decision, not a quarterly meeting between departments.
October 2024: ChatGPT search widens the citation contest
Search visibility stops meaning only Google page one
OpenAI launched ChatGPT search on October 31, 2024, adding live web search and inline citations for Plus and Team subscribers before a broader rollout in December. Perplexity was already answering prompts directly rather than presenting a classic results page.
This was a separate turning point from ads entering AI Overviews. Google had made its answer a place for citations and paid units; conversational engines showed that a buyer might ask for a comparison without visiting Google’s results page at all. A prompt comparing enterprise SOC 2 compliance platforms for a 50-person engineering team is not the same reporting unit as a short keyword. The engine retrieves information, forms an answer, and shows a small set of sources. The buyer may never see your page-one ranking.
That does not mean every search moved to a chatbot overnight. It means a Google-only position report became an incomplete view of discovery. If a competitor supplies the facts these engines use, it can appear in the answer while your carefully ranked page remains out of sight. Understanding why ChatGPT cites a competitor became a practical question, not a side project for whoever owns the AI slide in the deck.
Check the answers buyers actually request across engines. A citation is worth tracking wherever the buyer encounters it, not just where your existing dashboard knows how to count it.
March 2025: AI Mode turns the result into a conversation
Follow-up questions change the unit of work
Google introduced AI Mode in March 2025, powered by Gemini 2.0. Instead of treating every search as a single query followed by a list of links, AI Mode let searchers ask follow-up questions, narrow parameters, and refine criteria in place.
That matters to both sides of a search program. An organic page may need to answer the comparison, constraint, and follow-up question rather than repeat one target phrase. On the paid side, advertisers cannot simply select an “AI Mode keyword” and bid on that isolated term. Eligibility and placement run through the automated campaigns Google uses to evaluate the broader context.
The Search Engine Land account of AI Mode ads reported sponsored text ads in 29.45% of responses across the commercial US queries it examined. It also described a shift from short keywords toward longer conversational searches: the impression share of one-to-two-word keywords fell from 42% to 24%. The practical problem is not that every short keyword stopped working. It is that a buyer can express intent through a sequence of questions that a short-keyword report does not explain.
A five-check audit for the changed result page
Do this on core commercial pages and the buyer prompts they are meant to answer. Earning a citation does not require a stack of speculative new articles; fixing a page already near position eight may put existing visibility to better use. Keep the audit short enough to repeat, and make paid and organic teams look at the same queries.
- Check crawler access: Confirm that search and answer engines you want to reach are not inadvertently blocked by
robots.txtor inherited firewall rules; do not treat every AI-related user agent as an indexing crawler. - Put the answer near the question: Add a direct, self-contained response beneath the relevant
H2, using a comparison table when that explains the choice faster than prose. - Check structured details: Keep product attributes, pricing, and company information consistent across relevant page content, JSON-LD, and directory listings.
- Track citations for conversational prompts: Record whether your domain appears when buyers ask multi-sentence questions in ChatGPT, Perplexity, and Google’s AI results, not just where a short keyword ranks.
- Compare paid presence with organic citations on the same queries: This is the check teams most often skip. Without it, they can bid aggressively in an answer where they already have an organic citation while neglecting queries where they have neither kind of presence. The cost is a budget allocated without the full visibility picture.
The checklist does not promise that a crawler will cite a page because its answer is concise or its schema is tidy. It makes the page legible, then checks whether the intended result happened. Measure the citation before calling the work done.
Early 2026: The answer slot ties organic and paid together
The next decision is an allocation decision
Read the entries in order. SGE gave sources a place inside a generated answer. AI Overviews brought that answer into ordinary US search, and citation research showed that the cited URLs did not simply mirror the top rankings. Google then brought ads onto the same answer surface. ChatGPT search widened the contest beyond Google, and AI Mode made the query itself more conversational.
The throughline is not that blue links vanished. It is that a ranked listing no longer describes all the search visibility you have, while an uncited listing can lose attention on pages where the answer resolves the query first. The reported click declines on AI Overview queries make the cost concrete. Ads give a business another route onto the answer surface, but each paid click has a price. The organic citation remains the seat that does not charge per click.
That is why I do not buy the old arrangement in which a paid team optimizes campaigns, an SEO team publishes articles, and they exchange spreadsheets once a quarter. One team can bid on a query where the other has already earned a citation. Both can miss the buyer’s actual question because each is reporting a different slice of it. What the deck calls “channel alignment,” I call checking the same result before spending more money on it.
This is the operating problem groas is built to tackle: specialized AI models execute bidding, budget allocation, search-query work, and organic visibility work continuously, while a named strategist owns direction and accountability against qualified pipeline and closed-won revenue. The point is not to replace judgment with a busier dashboard. It is to connect the paid decision to what is happening in the answer, rather than paying for inefficient hours spent reconciling reports afterward.
The line from 2023 to now points toward more decisions made inside answers, not a return to a tidy page of ten equal links. I would start with the buyer questions that matter, inspect who is cited and who is paying to appear, then change the page or the bid accordingly. Waiting was cheap when the answer was an opt-in experiment. It is expensive when the answer is where the buyer looks first.