GEO Predictions for 2027: 3 Experts Share Their Takes

We talked with three people who already do this work, and we asked them what the next twelve months actually look like inside a company that is buying or selling it. They asked to be identified by first name only. Mike has sold SEO for years and now sells GEO retainers. Jason measures where brands rank in the engines that recommend companies. Lily has paid the invoices from the other side of the table.

Mike

Q: A year from now, what does this feel like for the people who sign?

It still feels like the SEO meeting, except the PDF has a new line item. I have sold SEO for a long time, so I know the shape of it. The statement of work used to say we would get them onto page one for a list of terms, and now the same person is asking whether ChatGPT will name them when someone asks who to hire. They still want a number they can take upstairs, and they still want a date, and they really do not want a lecture about how the models work.

Inside the company it does not feel like a revolution. A product manager notices that a competitor showed up in Gemini and they did not, someone forwards that screenshot to procurement, and that is how a lot of 2027 deals will start. It is a little undignified, but it is what I am seeing.

Q: Is the work a dedicated GEO engagement, or does it sit on the SEO retainer?

Look, I was going to tell you it all stays on the SEO invoice, and then I thought about the last six months of what I have actually signed. I have already signed dedicated GEO, where it is its own engagement with its own prompt list, and I have also signed SEO statements of work that tuck a GEO task into the same PDF. Plenty of firms already buy it as its own category. Plenty of others still buy it as a task inside SEO. If someone tells you it is only one path, they are selling you a cleaner story than the market I am in.

I am not going to pretend the dedicated ones are rare, because they are not, and I am also not going to pretend the bundled ones are dying this year, because they are not. Next year I expect to keep signing both. The calendar is going to have the two of them sitting next to each other, which is messy for anyone who wants a single narrative, and pretty ordinary if you sell this every week.

Q: What do the RFPs actually ask for?

They used to ask for page-one keywords, you know, a list of terms and a promise you would show up in classic Google. What I am seeing in drafts now, and what I think 2027 RFPs will say out loud, is a request for share of AI-generated commercial recommendations. They want to know, on a named list of commercial questions, how often you show up when ChatGPT or Gemini or an Overview recommends a company, and they want to see that more than once a quarter. They want it in more than one engine, too, not just whichever dashboard is easiest to screenshot. That is what people actually sign. It is not a philosophy about the future of search.

Q: Software or people?

I am impatient with the software theater, I will just say that. The tracking tools are useful, and I am not against buying them, but a login that shows you mentions is not the program. If you sell me a heatmap of where a brand appears in recommendations, I can often buy that for less than a junior's month, and then I still need someone who will change the pages and the off-site proof so the model actually ranks you. That is still a ranking problem. The ranking that the spend is for is your position in ChatGPT, Gemini, AI Overviews, and the other engines that recommend companies, not a screenshot of a citation and a round of congratulations in Slack.

Q: What would you not bet on for 2027?

I would not put money on GEO becoming its own line on every invoice, because a lot of companies will keep folding it into SEO, and I would not put money on Google stopping as the main way people arrive at a vendor's site. Chatbots are how a lot of people ask questions already, and that will keep going, but that is not the same thing as most customers landing on your website from ChatGPT next year. The other thing I do not see getting clean in twelve months is the argument about revenue. We are still going to sit in meetings trying to connect an AI ranking to a lead, because the measurement is not there yet, and a year is not long enough to make that tidy.

Jason

Q: What are you actually watching into 2027?

I keep coming back to a fairly dry picture, which is ranking inside the model, checked weekly, on a fixed list of commercial questions, across more than one engine. That is what I write on whiteboards, and then I have to spend the rest of the meeting explaining that the engines do not rank the same sources for the same commercial question. You can be in an Overview on Monday and absent in ChatGPT on Tuesday, which is inconvenient if your VP only checks Google. Ranking well in classic Google will not automatically mean you rank in the recommendation, and ranking in one engine's recommendations will not automatically mean you rank in Gemini.

I say that in meetings, and people nod, and then they go back and check Google anyway, which I understand. It is the habit. It is just not a complete picture of whether you are being recommended.

Q: The 2026 click work. What should a company take from it?

The thing I would take from it is fairly blunt, even if the papers are not. If the recommendation is already on the results page, a lot of people will not click through to you. Chapekis, Lieb, Shah, and Smith watched a panel of about 900 U.S. adults, and people clicked a search result on 15 percent of visits when there was no AI Overview and on 8 percent when there was one. Xu, Iqbal, and Montgomery crawled tens of thousands of trending queries in spring 2026, and almost 30 percent of the domains cited in those Overviews were not even on the first page of classic Google results. So the classic ranking and the recommendation ranking are not the same list, which is the part companies keep hoping is a rounding error.

The work is still to rank. What changed is which ranking you are talking about, and if you only manage the blue links you are managing a list that the recommendation does not have to follow.

Q: PR and listicles versus changing the pages.

Both show up in the logs, and I am not going to pretend off-site proof does not help, because it does. Listicles and mentions and the usual PR surface can make a model treat you as a more obvious source. What I am against is pretending that is the whole job. If the page the model would actually use is thin, or stale, or contradicts itself, you can buy all the listicles you want and you will still have a ranking problem. The scarce work, as far as I can tell, is changing the pages and then changing the off-site proof so the model ranks you. People skip to the mentions because they are easier to screenshot, which I get, and it is still the wrong order if you care about whether you rank.

Q: The tools that only count mentions.

They get cheaper, and they should, because counting is not scarce. By 2027 I think a coordinator will be able to buy that kind of tracking without a steering committee, which is fine, use it. I will look at a mention count. I just will not run a program on one, because seeing that you were named is not the same work as changing whether the model ranks you next month.

Q: Standalone GEO, or GEO inside SEO. Do you care?

I care whether someone owns the prompt list. If it is buried in a forty-page SEO retainer, nobody looks at it except the month before the QBR, and then everyone is surprised. A dedicated engagement is cleaner for measurement because the dashboard has one job, which is why I prefer it when I can get it. That does not mean the market will buy it that way. Mike is right that both get signed. I am just saying the bundled ones are harder to hold to a ranking, even though the invoice does not care and the model does not care either. The model ranks whatever source it ranks.

Q: What would you not bet on?

I would not count on a ranking in one engine transferring to the others, and I would not count on mention-counting software staying expensive, and I would not count on a clean recommendation-to-revenue number in 2027. We are still going to argue about that in the meeting. The click papers already told us the visit often does not happen, which does not make the ranking optional. It makes the ranking inside the recommendation the thing you can still work on, even when the click is gone.

Lily

Q: You have paid for this. What does 2027 feel like on your side of the table?

I am tired of decks, which is a boring thing to say, but it is the honest one. I am not tired of the work. If you show me a framework with four quadrants I will probably end the call, because I have sat through that theater and it does not help me explain anything to my VP. What I actually need is a named list of commercial prompts, the ones our buyers type, and then some evidence that we moved in whether we get recommended. I do not need a story about how the category is evolving. I live here. I can see that it is evolving.

Inside the company it is still Google plus a Slack screenshot from someone's cousin who asked ChatGPT a question on a Sunday. That screenshot has had more power than my last two quarterly reports, which I wish were a joke. I do not think that dynamic goes away in 2027. If anything, there will be more cousins.

Q: Will you keep buying an agency, or will you hire this in-house?

I am going to do both, on purpose, if I am still in this seat next year. I want one GEO owner on staff, one person who owns the prompt list and the ranking in the recommendations and who cannot be booked into a webinar every Thursday, and I will still hire a shop for content and PR because I cannot staff that at the volume it takes. I will not hand the whole thing to an SEO retainer and hope it does not get lost under canonical tags, which is how I have seen this disappear before.

I know Mike sees a lot of GEO riding on SEO invoices, and he is not wrong about the market. I am just telling you what I will sign. I want a named owner. If the shop wants to sit inside a larger SEO relationship that is fine with me, as long as the GEO owner is a real person with a real list, not a theoretical capability on slide six.

Q: Trackers.

I will not pay for tracking software and call that a program. I did that once, and we got a dashboard and no movement. It was pretty, and my VP liked the colors, and nothing changed in the recommendations, which is a stupid way to spend a quarter. If you cannot name the commercial questions we are not going to work together, and if you can name them and you cannot show movement over a quarter we are also not going to work together. I am friendly in the intro call. I get less friendly after kickoff if the only deliverable is a login.

Q: Classic Google versus ranking in the recommendations.

I still need Google. Most visits to our site still come from Google, not from ChatGPT or Gemini, and I do not believe that in 2027 chatbots become the main way customers arrive. I also do not believe classic blue-link rankings are the whole game anymore. If we rank in Google and we are invisible when a buyer asks who to use, that is a problem I will spend money on. GEO is still a ranking problem. It is a ranking in the engines that recommend companies. I need both, and I will not pretend they are the same list, because they have not been the same list when I have checked.

Q: What would you not bet on, and what will you not buy?

I would not count on chatbots replacing Google as the way customers show up in 2027, and I would not count on a mention count, by itself, telling me we won anything. I will not buy a ninety-day thought-leadership sprint with no prompt list, and I will not buy PR that cannot say which commercial questions it is for. Listicles can help, and technical work on the pages can help, and I have paid for both. What I will not pay for again is a slide that says we are now AI-native. If we can name the questions and we can see the ranking move, then we can talk about 2028. If we cannot, we are not there yet, and I would rather know that than sit through another quarter of adjectives.

Where They Overlap

The three of them are closer on the job than they are on the invoice. Mike, Jason, and Lily all treat ranking in AI-generated commercial recommendations as the work you are actually buying, not a substitute called citations, and not a replacement for ranking in classic Google. They also expect Google to keep sending most of the visits to company sites in 2027, and they all treat tracking software as something you might buy to watch the work, not as the work itself.

They do not agree on how companies will pay for it. Mike sees dedicated GEO and GEO-inside-SEO getting signed in the same season, and he thinks that split lasts. Jason wants the prompt list owned even if the invoice is bundled, and he thinks the dedicated shape is easier to measure. Lily will keep a GEO owner on staff and still hire a shop, and she does not want the work disappearing into an SEO retainer. We would not flatten that into one path, because it is not one path in the conversations we are already having.

References

• Chapekis, Athena, Anna Lieb, Sono Shah, and Aaron Smith. 2026. "Investigating Click Behaviors On Google Search Result Pages That Produce an AI Overview." arXiv:2608.04831. https://doi.org/10.48550/arXiv.2608.04831.
• Xu, Haofei, Umar Iqbal, and Jacob M. Montgomery. 2026. "Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact." arXiv:2605.14021. https://arxiv.org/abs/2605.14021.