FIR #479: Hacking AI Optimization vs. Doing the Hard Work

Sep 1, 2025

Ask episode

Chapters

Transcript

Episode notes

Posts and videos featuring Generative Engine Optimization (GEO) hacks and formulas are flooding the web. We reported recently on one such hack focusing on press releases. But when you consider the kind of content on which the AI models rely for their answers, it may be more efficient to revert to good, old-fashioned PR and marketing.

Links from this episode:

The next monthly, long-form episode of FIR will drop on Monday, September 29.

We host a Communicators Zoom Chat most Thursdays at 1 p.m. ET. To obtain the credentials needed to participate, contact Shel or Neville directly, request them in our Facebook group, or email fircomments@gmail.com.

Special thanks to Jay Moonah for the opening and closing music.

You can find the stories from which Shel’s FIR content is selected at Shel’s Link Blog. Shel has started a metaverse-focused Flipboard magazine. You can catch up with both co-hosts on Neville’s blog and Shel’s blog.

Disclaimer: The opinions expressed in this podcast are Shel’s and Neville’s and do not reflect the views of their employers and/or clients.

Raw Transcript:

Hi everyone, and welcome to For Immediate Release. This is episode 479. I’m Neville Hobson.

And I’m Shel Holtz on Thursday Neville, you and I are going to interview Stephanie Grover, who is the marketing and PR director at Horowitz Agency. This is a marketing agency that works with law firms, production companies, and other professional service providers in the US and Canada.

And we’re going to talk, be talking about GEO generative. Optimization, generative engine optimization. I’m not altogether sure, but it’s a hot topic and I thought I would take today’s episode to set the stage for that because we’ve all seen the headlines recently. Chat, GPT, traffic referrals to websites plummeted more than 50% in a single month.

This summer, and that’s not a blip. It’s a structural change in how these large language models are surfacing content. [00:01:00] OpenAI tweaked its ranking and suddenly chat GPT Beca became began citing fewer sources, leaning more heavily on places like Wikipedia and Reddit. Useful for users. Yeah. But if you’re a brand counting on visibility, it’s a gut punch.

And meanwhile, the volume of content keeps exploding. A new B2B study found content production has tripled year over year, which could be partly attributable marketers flooding the zone with content in the hopes. LLMs will hoover it up and they’ll show up in AI search results. Interestingly, that tripling of content volume has not been accompanied by commensurate budget increases.

Mm-hmm. But we’re producing more content than ever but it’s not necessarily better content or content that LLMs are actually going to use. So no surprise that there’s a scramble for the supposed hack that will unlock, sorry. Unlock. Okay. Unlock [00:02:00] rhymes with hack. What can I say? So no surprise that there’s a scramble for the supposed hack that will unlock generative engine optimization.

GEO. Some companies are starting to figure out that it’s not about gaming the algorithm, though. It’s about trust. Sylvia la this chief marketing officer at Kenji shared a fascinating case study on LinkedIn. Her team created the sequence, it’s a standalone content brand with its own domain separate from the corporate site.

The idea was simple. Create a community driven media hub. Human high quality, free of fluff. The unexpected bonus that came from this is that LLM started treating the sequence as an external authority. When asked about can g chat, GPT doesn’t just reference the company, it references the sequence. In other words, by building a trusted resource that stands on its own apart from the central [00:03:00] brand site, they built credibility, not just with their human audience.

But with the algorithms too that aligns perfectly with something else I saw from Liza Adams, another CMO. Who pointed out that the reason Wikipedia and Reddit dominate AI citations isn’t mysterious, it’s because they directly answer real questions using the same plain language. Real people use Adams contrasts two types of marketing teams, the ones who do the hard work of auditing their content, listening to customer language, and creating genuinely helpful answers.

The ones chasing quick fixes and shortcuts. Her takeaway is that there is no algorithm hack AI amplifies what’s already there. If your content is genuinely useful, trustworthy, and present in the watering holes your customers rely on, the algorithms will pick it up. If it’s not, there’s no trick. That’ll save you.

Now, add one more layer to this. What people themselves actually trust according [00:04:00] to new LinkedIn research networks, our peers, our colleagues, people we know still rank as the number one trusted source of information far ahead of. AI searches or even traditional search engines. In fact, 43% of professionals say their network is their first stop when they need advice at work and new.

Nearly two thirds say colleagues help them make decisions more confidently Now. Think about that for a second while marketers obsess over how to get cited by chat, EPT, or Gemini, the real influence continues to live in trusted human connections. That’s also changing how brands approach content. LinkedIn reports that 80% of B2B marketers are increasing investment in community driven content.

Bringing in creators, employees, and subject matter experts, they understand that credibility doesn’t come from corporate channels alone. It comes from trusted voices. People want to hear from. So where’s that leave us in? A [00:05:00] messy transition is where that leaves us. Generative AI referrals are volatile.

Content volume is ballooning. Everyone’s chasing GEO or a EO or A-I-S-E-O or whatever acronym you want to use, but the truth is the winning strategies are old school in the best sense. Build trust, answer real questions, and put your content where your community is already paying attention. Kenji’s story shows that if you can create a true content brand, it can earn authority with humans and with machines.

Liza Adams reminds us that shortcuts don’t work answering real questions and real language does. And LinkedIn’s research makes it clear that even in the age of AI are human networks still shape the bulk of decisions. So stop chasing algorithms, earn trust. That’s the real path to visibility in a generative search world.

Good. Good assessment there. Shell, I [00:06:00] think we’ve talked about this a lot on recent episodes of this podcast. One that comes to mind recently, which is in this kind of area, was the episode in which we talked about the, new form of press release that some people are discussing that is aimed at algorithms, not humans.

And this to me is in the same kind of territory in that context. It’s a hack. So you, yeah, you call it that if you like, but I see it more as this is the direction of travel and the critics who say, no, no, no, no, no. This is not gonna take off it. Trust me. It really is. We’re seeing signs everywhere that this is, a really good sorry, not Google. It was Pew Research at the end of July. Had a quick survey results. Again, we’ve talked about this stuff on this podcast. Google users are less likely to click on links when an AI summary appears in the results. That’s been a kind of recurring topic, what a communicator’s gonna do about that.

It’s not about, well, it’s not what? It’s [00:07:00] not about. What it is about is making your content discoverable more than it currently is in light of this trend that is happening. That means you’ve gotta rethink your approach to such and that kind of acronym soup that you reeled off just now. She maybe think of, oh, McDonald had a farm, right?

E-I-E-I-O, right? You remember that? Search engine optimization and generative engine optimization seems to be the latter seems to be the common descriptor of all that I have seen particularly when I was looking into this area. But the reality is that. Change is underway. So, SEO if I contrast it this way, SEO’s the purpose of SEO, if you like, is to improve a website so that it ranks higher in search engine results.

That’s essentially what SEO’s all about. GEO on the other hand, is this new idea, and as you mentioned it shaped by the rise of generative AI tools like Chat, GPT, et cetera, that instead of optimizing for search engines and websites, it’s. [00:08:00] GEO is about making content more discoverable and usable by AI systems that generate answers now.

That to me, therein lies exactly the threat, if you will, if you wanna see it as a threat. I see it as evolution, and I see it as opportunity for those looking for. Angles looking for leverage, looking for an opening that you know, if your content is gonna be more discoverable and usable by AI systems that generate answers, you are on a winning streak in that case, because this is the definite trend.

So it means you’ve gotta structure information ways AI models can easily interpret and cite. Almost that same phrase was used in that topic I mentioned that we discussed on a previous episode about the kind of evolutionary press release that’s emerging for the AI era to coin to nearly coin an already coined phrase.

This is, in my view, this is the trend that’s going to knock the existing model sideways. And if you are not. [00:09:00] Looking into how you need to be part of this. You’ve got hard work ahead if you are able to catch up with people who are doing that. Key differences this is quite simple. SEO attracts people by clicking on links.

That’s the point of it. GEO AI engines to ensure your content is included in the answers it generates. Think about that. And that’s outside your control in the sense of the generation, but it’s totally within your control to make your content discoverable. So that is included in those generated answers.

So the format SEO uses see a rank list of websites. GEO people see a synthesized. Without clicking through to the original site that’s, you’re not gonna change that. It’s interesting, there are many similarities and this is this is a kind of paradox. Both aim to increase visibility of your content.

Both rely on credibility. Both evolve constantly as the technology changes and both can affect reputation. If you’re doing badly, like keyword stuffing for SEO [00:10:00] or gaming AI for GEO, you risk undermining trust, and that’s the underlying point of it all. So. I think I’m looking forward to this discussion with Stephanie on, on Thursday, by the way, because I’ve seen when I did a little quick lookup knowing we’re gonna talk Stephanie on, on what she was gonna talk to us about.

I see a lot of criticism of geo. Some people call it snake oil. This is not genuine, it’s not authentic. ’cause there’s no, this is definitely not what you should be doing. I’m thinking deja vu in the sense of what I’ve seen in the past when something new appears in the PR realm, if you like. I think it is something definitely to pay attention to.

I think what you mentioned and the links that will include in the show notes indicates that SEO is not gonna suddenly drop dead. Not at all. But the future is not. Links on websites at all? Not really. I don’t believe. I can see, we’ve talked about this again on the previous episodes about a IO reviews from Google that I see conflicting stories, and I think you might have referenced something in [00:11:00] the beginning shell that trafficked on in that era is diminishing recently according to sub reports.

But the truth of the matter is that, that ask almost anyone, there are surveys saying this, people prefer that. Users, let’s call them that, right? Customers or whoever you call looking for something, there’s a result. You see what you need in that little box on the right hand side of your screen. And that’s as far as you go.

You don’t click on anything necessarily. And that’s people’s preferences. Why are you gonna go to a website when hist history shows you that it’s not? Much you can trust. So you’re looking, for instance, and I’ve used this example before. You’re looking for electric cars ’cause that’s a hot topic.

And here in the UK the government just announced a rebate. To stimulate people buying electric cars. So you go online to look at what are the deals in electric cars who’s got this particular model. And as you do a traditional search the chances are high that amongst the results you’re gonna get in that list are gonna be [00:12:00] car dealers and sponsors.

There’s all, that word appears everywhere. I wouldn’t go there at all, not knowing that there’s now this. Far more trustworthy alternative that’s likely to give me what I want without the feeling that you’re being manipulated by an algorithm because they’ve paid money to have their listing appear.

This is where we’re at in the territory it seems to me show. So I, again, Stephanie, looking forward to hearing how can she put this forward as a good explainer for communicators? And this is the message we mentioned before, is that, you’ve got about, you’ve gotta really make sure your content is clear, credible, and discoverable.

That’s not a new thing. You’ve always had to do that, right? But now it’s becoming more likely that if it’s not, you’ll be ignored, and that’s not good if you’re a brand. No. And you said SEO’s not dead. And it’s not gonna die anytime soon. Maybe it’s, well, it’s received a diagnosis and it’s physicians are trying to figure out if there’s anything they can do [00:13:00] to save it, but it’s has time yet to enjoy what’s left of its life.

They I’ve heard people talking about the notion that one day the web is going to be designed for bots that is going to be bots, that are going to be visiting websites and gathering information and fueling the research and the searching. That humans do? I don’t think that’s accurate.

I think there’s always going to be reasons people want to visit websites. I mean, is the Onion website really just there for bots to hoover up? Its content. No. People wanna go have a laugh. People wanna go watch videos on YouTube. There are reasons that people are going to want to go visit some kinds of websites.

But, look at that. That item that we reported on about a new type of press release to accommodate the algorithms. And the more I think about that, the more I believe that if you need to revise the approach you take to press [00:14:00] releases to accommodate large language models, then the old approach you were taking to press releases probably sucked because what is it that appeals.

To the large language models well, first of all, they’re not sharing their algorithms with us. When Google makes a change to Google, they write papers about it. They give the new update a name and all the SEO companies gobble up this information and make s. Changes to their work in order to accommodate the changes that Google’s made.

What? OpenAI and Google Gemini, Google DeepMind, and the other players in this space what their algorithms are is a black box. We don’t know. So if you’re trying to game the algorithm let’s face it, you’re guessing. What we do know is roughly how. These models work not in any great detail.

Even the people [00:15:00] who built them don’t understand how they work in any great detail. But we know that the knowledge that is collected in the training and in the continuous training that goes on through adding the new stuff that is posted, that the bots are all out gathering is all stored in cells and neuro net.

In neurons in a neural network. And it’s the association between the neurons, just like in our own brains that produce the next token, the next word, the next sentence, the next paragraph in the results that you see. And we do know that a lot of these results are coming from Reddit and from Wikipedia, and we do know, or at least it is a.

Very strong strongly supported, bit of surmising that it’s because they talk in plain language and answer questions that people wanna know about. So if you if you consider that website that [00:16:00] the brand Cange came up with that was a content driven site separate from the brand that was community driven.

It worked. It satisfied both the need to build trust and interest among your community. And the LLMs liked it and started citing it when people asked about Cange. So it seems to me that rather than try to do something different, the only thing different you should be doing is if you’ve been doing it wrong up till now.

You should go get yourself a good primer on marketing and a good primer on public relations from maybe the last 10 years and read that and do that because that’s what the l lms, like, I don’t think there is a magic bullet. I think that what we are good at when we do it the best that it [00:17:00] can be done.

That’s what works. And that’s the direction we need to head. I agree. I totally agree. I think part of the problem though, for communicator certainly is the how bit, how you can do that. So, so one question I have, given the shift that’s underway, it is SEO is about ranking in a list of links.

GEO is about surfacing in AI generated summaries where the links may be secondary. So for communicators, how do communicators adapt when fewer people click to the source? No matter what the reasons might be, that’s the question, how do you do this? So it’s a how we need to help them, help ’em understand. And I don’t have the answer to the how myself either.

Well, I think the answer goes back to basic strategic planning, right? It’s Sure. What is your goal in the first place, and what strategy are you going to employ in order to achieve that? Goal. And if the old strategy of SEO doesn’t work, if the old even older strategy of having a website doesn’t do the trick anymore, what else do you do?

[00:18:00] You know, back to the drawing board. But that’s what we do. It is true and we’ve mentioned this before, so for instance uh, you know, AI models are more likely to size sources with consistent metadata, explicit effects breach. That’s common sense, I think. But publishers and brands worry. If AI provides the answer directly, what incentive is there to visit the source?

So this is a circular point. Go back to the other one. How to communicate adapt when fewer people click through to the source. Uh, like all things, you need a plan, but what is certainly clear, and I think that point I just mentioned, that would undermine ad advertising based revenue models, right? It doesn’t mean that next week is all gonna collapse.

Not at all. But that’s. You can see the writing on the wall. I’m sure I can certainly see writing on the wall about this, and I’m kind of a bit nonplussed by deniers. May maybe the, the VX people too. Deny all that. I, I don’t know, but I mean, it just seems to me crazy. You can see the changes right in front of your eyes, whether you like it or not.

So I think you mentioned [00:19:00] your gaming, uh, as I mentioned in my early combat gaming. You know, gave me the content for LLMs you know, AI friendly, but misleading content. I think that is still a, a big issue to be concerned about given how that is already going on. And, uh, we already, you know, we already hear the kind of worry signs about and these are the things that are in the realm of, you know, what happens if the robots do take over all gonna die, et cetera.

But it means that if. AI generated content is prompted by people with bad in. Now, these, these famous bad actors we hear about all the time, and the output is based on that bad input. And then others reference that over time. Uh, what is that giving us? Bullshit. Basically misinformation, disinformation how can we trust anything?

And we’re already seeing this demonstrated, and we talked about this in a recent episode in Wikipedia, where they are battling the [00:20:00] reality. Of the influx of misinformation and disinformation in content written by others that is faster to appear than the moderation system can handle it. So they’re now coming out with a quick delete methodology that is the short-term solution.

I dunno what the long-term one is for Wikipedia, but that directly impacts trust. And I think if, if that happens, then we’re in real trouble. But we’re seeing this, I, I see this daily here. The, the BBC is very much at the front. Page on this and a number of other media companies in the uk the verification team there.

There’s, I gather, there’s now like a hundred people on this team. I’m not sure where I, where I saw that, but, uh, it may not be true in that case, but it’s a lot of people who literally analyze everything. Uh, there was a, a story I saw this morning analyzing all this stuff over the weekend. I’m sure you saw it about.

Has Trump died? Is he still around? Is he still alive? Uh, and that’s been looked into to get to the source of where that started. I’ve not seen the [00:21:00] conclusion of that research, but Wikipedia’s definitely in the frame. Reddit I saw mentioned earlier, and, uh, every time I go to Reddit and I use Reddit a lot ’cause of things I’m interested in.

Uh, and it’s full of, opinion. And then the counter opinion, and then you get rabbit holes everywhere about, well, I heard it was this and that’s not true. Worse is the ones who say, with authorities of sounding words, that X it is so and so without any verification. People believe that stuff there is part of the problem too.

Humans are too trustworthy in that regard. We’re, it’s a mess. And how do we, again, how do communicators adapt when fewer people click through? The source is a kind of umbrella question to it all. Uh, it’s a difficult one. Well, I think the fact that fewer people are clicking to a source means that the, the purpose of the source needs to change.

You need it there because that’s what the LLMs grab in, in, in their training sets. So you. [00:22:00] Want to continually be working on that content. And there may come a day where you won’t care if a single human sees it, as long as it’s delivering the AI results that you need it to. Perhaps website design is going to diminish only the sites that people actually do want to visit for whatever reason will get high quality design.

The rest will be, you know, mark more markup language. So, uh. That it’s even easier for the LLMs to grab. I don’t know where that’s headed. We’ll see. No, uh, because I’m sure that that will follow whatever ends up working. But the idea that you won’t need a website anymore is absurd if you are generating revenue off of that website through advertising.

You’re probably gonna need to switch to something more community driven. Right? Because people will continue to go to community sites. Yeah. Uh, because a AI can’t replicate that. They can’t deliver that for you. No, [00:23:00] I, I, I agree. And I don’t see websites dying away, even if, uh, some people say they will. ’cause like you said, people will always want to do certain things that might well diminish nevertheless.

But in all of that, you mentioned the point earlier, and you’ve mentioned this the other day, I think in, in, uh, in some of your recent experiences that, already there are tools that, uh, let’s call them AI agents, even though I don’t think they really are yet that will go out to YouTube and find these, take these 17 videos on this topic that you are really interested in.

Summarize it all from the, uh, from the transcript and present to you a summary of what it’s all about with some actionable items for you to do, depending on what you’ve asked it to do for you. So that to me, I, I can see more likely I speak of myself here, uh, and I, I’ve kind of. Scroll through YouTube looking for stuff myself, that would be far more appealing than me going to the website.

Well, I’ve actually created a, a, a tool that does, so I wouldn’t, I wouldn’t, I wouldn’t, I wouldn’t do that. And I, I, I would imagine a lot of people would prefer to do that than go themselves to [00:24:00] the onions. An example you made, have someone give you the list of, this is what this is all about. I’ve done that.

I’ve, uh, used, uh, Google’s opal tool, uh, to create a workflow where I enter the URL of a YouTube video. I say, this is a video that gives instructions on how to do x go to the transcript and just give me the steps to, to do it. Yeah. I, I. I chose that video because I watched it and it has the steps I want.

What I don’t wanna have to do is continually rewind, uh, as I’m going through these steps. So once I’ve seen it, I just wanna see the steps articulated. And it works real well. I used it a few times. Then I got the comment browser from Perplexity, which I can just do that as a prompt. Now, uh, I, I don’t need a workflow, so I, I just got that over the weekend as my, you know, congratulations.

You’re now in the trial, blah, blah. I’ve installed it, but I’ve not, uh, tried it yet, but it’s on my to-do list to do. I see like you watch a couple of YouTube videos [00:25:00] to see everything you can do with it, because there’re, I’m gonna try, you’re not obvious. Try. I’ll try it with that. I, I suspect so, so this adds to the, to what’s happening in the development phase of what’s going on in that wide world out there.

Uh uh, and the pressure is definitely on to find. Better ways, more effective ways to, uh, to get the answers to what you are looking for that are far more, uh, uh, appealing to users. This is really what it comes about. Yeah. What do you prefer to use? And you will ignore lists of links in the future ’cause they’re not.

Sufficient information for you without more work. So I find a development in, in tools like chat, GBT with the recent models where they now do what Perplexity started doing quite a while ago, which is giving you the list of sources. So I did something the other on a research project that I’m working on right at the moment that I, I, I told chat GBT to give the full list of sources, [00:26:00] which it did.

Citation linked to the appropriate passages. Made life a lot easier. A lot easier. Not as good as perplexity, but it’s getting there and didn’t do it until not long ago. Plexity, I, I use occasionally, but chat gpt is my go-to for all these kinds of things. Uh, particularly model five. Initially skeptical that as I was, but the thinking mode is really quite good.

A bit slow though, I have found. So, um, this is part of the, the deal. Right. Again, like I said, I’m looking forward to this conversation in the FIA interview we’ve got scheduled for Thursday. Might, uh, might enlighten us a bit more. Looking forward to it myself and until then, that’ll be a 30 for this episode of four immediate release.

The post FIR #479: Hacking AI Optimization vs. Doing the Hard Work appeared first on FIR Podcast Network.