No. SEO optimizes for a position in a list of links. AI Optimization changes what an engine believes about your business so it names you in an answer. They overlap — technical SEO is one of the levers we use, and good SEO content is often what gets cited — but the goal, the measurement, and the starting point are different. AI Optimization starts with reputation and sentiment. SEO does not.

Because AI Mode is becoming Google search. Every local-intent query we tested surfaced AI Mode, and it cites Google Maps in 79.8% of citations — not the ranked pages your SEO earned. Position one already loses about 58% of its clickthrough when an AI Overview sits above it. Strong SEO is a genuine advantage going into this work. It is not a substitute for it, and it is not protection against the surface changing underneath it.

Because the engines are reading different sources. In our research, Gemini and ChatGPT shared only 8.3% of cited domains and named the same top business on 4.2% of queries. Even within a single engine, repeat queries returned the same top business just 7.9% of the time for Gemini — against 90.2% for Google’s local pack. Inconsistency is the normal state of these systems, not a sign something is broken on your end.

No, and be careful with anyone who does. Answer engines are not stable enough to guarantee a position in — our own drift research is the evidence for that. What we commit to is the work, a consistent cadence, and reporting that shows what moved, what reverted, and what we did next.

No one does, and we will not claim otherwise. There is no keyword volume tool for private conversations with a model. We build informed query sets from your industry and services, mine People Also Ask and related-question data, and research directly in the engines and against their APIs. It is a strong process and it is ours — but it is inference, and we label it that way in your reporting.

It depends entirely on which surface you mean, and for local businesses that distinction is expensive to get wrong. We crawled 25,898 cited pages. Schema was present on 64.0% of pages cited by AI Overviews but only 58.7% of pages cited by AI Mode. Schema tracks with citation on AI Overviews and considerably less on AI Mode — because AI Mode, on local queries, is assembling answers from Google’s pre-packaged local data rather than from your page.

Correlation is also not causation. Ahrefs tracked 1,885 pages that added JSON-LD and measured no significant citation lift afterwards. Pages that already deserve citation tend to carry schema; adding schema to a page nothing else recommends does not produce one.

Practically: LocalBusiness and FAQ schema is worth implementing and we do it properly. It is not what wins the AI Mode answer for a local business — your Google Business Profile is. Be wary of any AEO package whose main deliverable is schema.

For local businesses it is the single most important asset. Google’s AI Mode sends 79.8% of its citations to Google Maps. That makes your profile the primary record the answer is assembled from — categories, services, attributes, hours, photos and the substance of your reviews are all AI Optimization work, not routine local SEO housekeeping.

There is a threshold at roughly 4.5 stars, and clearing it matters far more than exceeding it. In our citation research, 90% of AI-recommended businesses were rated 4.5★ or higher and 97% were above 4.0★. But the businesses AI recommended averaged 4.75★ against a plain-search baseline of 4.84★ — the baseline was the higher-rated group. Within the recommended set, rating correlated with how often a business got named at 0.013, essentially zero.

So reputation is a pass/fail gate, not a dial you can turn. Get above about 4.5★ and stay there; pushing from 4.7★ to 4.9★ buys you almost nothing from the engines. Volume behaves the same way — AI-recommended businesses had a lower median review count than the baseline (295 vs 322), with the average pulled up by a handful of large outliers.

What moves things is what the reviews say. Answer engines read them rather than counting them: they weigh depth over volume, detect repeated phrasing that signals manipulation, and pick up mixed sentiment buried inside a high star rating. The consumer gate is tightening too — 31% will now only use a business rated 4.5★+, up from 17% a year ago, and 97% of AI users double-check a recommendation against real reviews.

We work on it regardless of whether it is accurate. That is deliberate — an engine acts on what it has absorbed, not on what is fair. A resolved complaint from three years ago still shapes the answer if nothing has displaced it. The levers are the same: engaging in the conversation, publishing content that answers it, technical SEO so that content is readable, and third-party coverage. Where the complaint turns out to be accurate and recurring, we will tell you that the honest fix is operational.

Usually weeks, not months — when the recommendations get implemented. Some changes take longer, and some results flip back as the underlying data shifts, which is why monitoring is continuous rather than a final phase. In one controlled Google Business Profile services test we measured movement within 72 hours that then held for 21+ months. Sentiment repair is slower, because you are waiting for new signal to accumulate and displace old signal.

A plain-language report of activities, dates, and results — written for a business owner, not a marketer. Where you provide Google Search Console and call tracking access, we tie the work directly to contact form submissions and call volume. Being recommended is where measurement starts, not where it ends: a citation that produces no calls is a vanity metric and we would rather find that out early.

Yes, and more than before. Your website is a citation surface, not just a landing spot — Gemini sends 59.9% of its citations to business websites and AI Overviews sends 73.5%. If a crawler cannot read your pages, no amount of reputation work will get them quoted. Technical SEO is one of the five levers we use to move sentiment, because corrective content that is not crawlable does not exist as far as an engine is concerned. What changes is the goal: SEO used to end at the ranking. Now the ranked page is raw material for an answer someone else assembles.

Make yourself the easiest business in your market to verify, then make sure what gets verified is positive. ChatGPT is not consulting a ranking — it is assembling an answer from whatever sources it grounded on for that particular phrasing, and those sources move. Ours moved dramatically on August 20, 2026, when ChatGPT’s Reddit citations for local queries dropped from 41.7% to 0.0% and its citations to business websites jumped from 15.9% to 42.8% overnight.

Practically that means: an accurate, complete entity record; review substance that a model can quote rather than just a star count; content that answers the actual question in extractable form; and third-party coverage on surfaces the engine is already reading in your vertical. There is no universal source list — in our data Reddit dominated auto repair while legal directories dominated personal injury. Which surfaces matter is a per-vertical question, and we test it rather than assume it.

They matter less than ChatGPT and Google today, and they behave differently enough that you cannot assume one strategy covers all of them. Published 2026 referral estimates put ChatGPT far in front, with Gemini, Perplexity, Copilot and Claude sharing the remainder. Their citation habits diverge sharply — Claude overlaps only about 13% of its cited domains with ChatGPT, and both Claude and Copilot lean heavily on LinkedIn for social citations where ChatGPT leaned on Reddit.

Our method is engine-agnostic on purpose. We sample across engines rather than optimizing for one, because the underlying work — accurate entity, positive sentiment, credible third-party coverage — is what all of them are reading. Chasing a single engine’s current quirk is how you end up with a strategy that expires on a Tuesday.

No. Ahrefs analyzed 137,000 sites and found 97% of llms.txt files received zero bot traffic — the crawlers simply are not requesting them. No major AI provider has committed to reading the file in production, and Google states plainly that you do not need to create machine-readable files or Markdown to appear in its generative features, and that it ignores them. John Mueller has compared llms.txt to the keywords meta tag.

It costs almost nothing to add, so we will not talk you out of one. But if it appears as a line item on an AI optimization proposal, ask what else is on there.

By sampling repeatedly instead of checking once. A single lookup is close to meaningless on a surface where repeat queries return the same top business 7.9% of the time. We run a defined query set across engines and across phrasings on a schedule, and report the pattern — how often you appear, for which question shapes, on which engines, and what is being said about you when you do.

Then we connect it to outcomes. With Google Search Console and call tracking access, your report ties the activity to contact form submissions and call volume. Visibility that never becomes a phone call is a number, not a result, and we would rather surface that early and change the work.

Ask four questions, and treat vague answers as answers.

What is your actual method? A real one names its steps — entity work, sentiment remediation, citation-surface mapping — rather than promising to “optimize your content for AI.” What will you report? Citation rates and AI-referred contacts, not impressions. How does this connect to my SEO? If they treat the two as separate products, they do not understand that the content being cited is largely the content that ranks. What can you prove? Ask for tested findings with sample sizes, not opinions about where search is going.

Four red flags worth walking away from: a guaranteed position in an AI answer (the surface is not stable enough to promise one); a deliverable that is mostly schema markup; a definitive report of what prompts your customers type, which nobody can produce; and visibility reporting with no line connecting it to leads.

Steady Demand’s AI Optimization starts at under $1,000 per month, against a published industry range of $3,000 to $50,000+ per month for comparable AEO/GEO retainers. It is built for established businesses doing $1M+ in annual revenue that already invest in SEO and other marketing — single-location and small multi-location businesses first, franchises and multi-location brands second, plus white-label programs for agencies.