What an AI SEO agency actually does
An AI SEO agency does two jobs at once. It uses AI tools inside the SEO workflow, for research, auditing, clustering and drafting, with people checking the output. It also optimizes your site so it can be found, understood and cited in AI-powered search: Google AI Overviews and AI Mode, Microsoft Copilot, ChatGPT search, Perplexity and Gemini.
The second job matters more, and it is not magic. Google says its generative AI features are "rooted in our core Search ranking and quality systems," so the foundations of classic SEO still apply: crawlable pages, clear content, strong evidence and a site people trust. What changes is the emphasis. You need passages that can be quoted and stand on their own. Entities and facts have to be unambiguous. Crawlers from several companies need access, not just Googlebot. And measurement has to look past rankings to impressions and citations in AI answers.
A good AI SEO agency is therefore an SEO agency first. It adds the AI layer on top of that work. It does not replace the fundamentals with it.
What Google and the AI platforms actually say
Many AI SEO services are sold with claims that do not match the platforms' own documentation. Before you compare agencies, it helps to know what the primary sources say. The table below checks common pitches against them.
| Common pitch | What the primary source says |
|---|---|
| "You need a special GEO/AEO strategy, completely separate from SEO." | Google: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" (Google Search Central, updated July 2026). |
| "Add an llms.txt file and AI engines will cite you." | llms.txt is a 2024 proposal by Jeremy Howard. Google says you don't need "new machine readable files, AI text files, markup, or Markdown" because Google Search "doesn't use them" (Google). Other platforms have not published equivalent commitments, so treat it as an optional experiment, not a ranking lever. |
| "We'll add special AI schema." | Google: structured data "isn't required for generative AI search, and there's no special schema.org markup you need to add" (Google). Accurate structured data is still good practice for eligible rich results. |
| "Rewrite everything into tiny chunks for AI." | Google: there is "no requirement to break your content into tiny pieces," since its systems understand multiple topics on one page (Google). |
| "There are extra requirements to appear in AI Overviews." | To be eligible as a supporting link, a page "must be indexed and eligible to be shown in Google Search with a snippet". No extra requirements exist (AI features and your website). |
| "We'll mass-produce AI pages to cover every query." | Using generative AI to create many pages without adding value may violate Google's scaled content abuse policy (Google guidance on generative AI content). |
| "We guarantee #1 rankings or AI citations." | Google: "No one can guarantee a #1 ranking on Google" (Do you need an SEO?). Nobody controls which sources an AI answer cites either. |
None of this makes AI search optimization pointless. It means the work that moves results is ordinary SEO done to a higher standard, plus a few AI-specific checks. It also means an agency that leads with shortcuts is a warning sign.
How AI search changes the work
Fewer clicks per search, so visibility inside the answer matters
Pew Research Center analyzed the browsing of 900 US adults in March 2025. Users clicked a traditional result on 8% of visits to pages with an AI summary, against 15% when no summary appeared. They clicked a link inside the summary itself on just 1% of visits, and 18% of the searches studied produced an AI summary (Pew Research Center, July 2025). That was one month of US data, so treat it as a direction, not a forecast for your site. The implication still holds. When an answer is shown, being one of its cited sources, and being the brand the answer names, carries more weight than an extra position in the blue links below it.
More than one crawler decides whether you can be found
In classic SEO, Googlebot access was most of the story. AI search adds crawlers from other companies, and each one has a different purpose. Blocking the wrong one can quietly remove you from an AI product.
| Crawler | Owner | What it is for | If you block it |
|---|---|---|---|
| Googlebot | Search index, which also feeds AI Overviews and AI Mode | Google can't crawl the content, so pages can't show a snippet or be used as supporting links in AI features (Google) | |
| OAI-SearchBot | OpenAI | Surfacing sites in ChatGPT search | Pages won't be shown in ChatGPT search answers, though they can still appear as navigational links (OpenAI) |
| GPTBot | OpenAI | Training foundation models | Signals content shouldn't be used for training. Separate from search visibility (OpenAI) |
| PerplexityBot | Perplexity | Indexing for Perplexity search results, not model training | Perplexity recommends allowing it to appear in results (Perplexity) |
A sensible AI SEO engagement documents your robots.txt and CDN or firewall rules bot by bot. The decision to allow or block each one stays with you. Training and search inclusion are different choices, and a good agency explains the trade-off instead of making it for you.
Measurement is finally possible, though still partial
Two first-party reports now exist. Google Search Console has a generative AI performance report. It shows how often links to your site appeared in AI Overviews and AI Mode, but it covers impressions only, with no clicks, CTR or position, and it reached all sites worldwide on 31 August 2026 (Search Console Help). Bing Webmaster Tools added an AI Performance report in public preview on 10 February 2026. It shows citations, cited pages and the "grounding queries" behind them across Copilot and Bing's AI summaries (Bing Webmaster Blog). For ChatGPT, Perplexity and Gemini there is no equivalent first-party citation report. Agencies rely on prompt sampling with third-party tools, which is useful but directional, not a census.
AI SEO services vs classic SEO: what an engagement includes
The fastest way to judge an AI SEO proposal is to ask what it adds to a normal SEO engagement and what it leaves unchanged. Here is how we scope it.
| Area | Classic SEO engagement | AI SEO engagement adds |
|---|---|---|
| Technical | Crawlability, indexing, speed, canonicals, sitemaps | Bot-by-bot access audit (OAI-SearchBot, PerplexityBot, Bingbot), snippet controls review (nosnippet, max-snippet), server-rendered text for key facts |
| Research | Keyword research, SERP analysis | Question and prompt mapping, AI Overview presence per query, which sources AI answers already cite in your niche |
| Content | Pages mapped to keywords and intent | Answer-first passages, clear definitions, cited facts, comparison tables that can be quoted accurately |
| Entities | Basic on-page optimization | Consistent brand, service and location facts across your site, profiles and structured data, so systems describe you correctly |
| Trust | Links and on-page signals | Clear authorship and "how this was made" notes (Google's Who/How/Why), sourced claims, digital PR that earns mentions as well as links |
| AI in the workflow | Manual or tool-assisted | AI-assisted research, clustering and QA with human review, never unreviewed mass publishing |
| Reporting | Rankings, clicks, conversions | Plus Search Console gen-AI impressions, Bing AI citations and a fixed prompt set sampled monthly |
Notice what stays the same: the technical base, useful content and genuine authority. Google itself states that E-E-A-T is not a single ranking factor and that "trust is most important" (Creating helpful content). AI SEO does not change that. It raises the cost of ignoring it.
An AI-search readiness checklist you can run yourself
Use this five-layer checklist before you hire anyone, then ask agencies how they would handle each layer. Work through it in order. A failure in an early layer makes the later ones irrelevant.
- Access. Can the crawlers that matter reach your key pages? Check robots.txt for Googlebot, Bingbot, OAI-SearchBot and PerplexityBot. Check CDN or firewall bot rules too, because these often block AI crawlers without anyone noticing. Confirm important text is in the HTML, not only rendered after user interaction.
- Eligibility. Are the pages indexed and allowed to show a snippet? A stray nosnippet or a very low max-snippet limits how Google can use the page in AI features (Google). Check Search Console page indexing for your money pages.
- Clarity. Does each key page answer its main question in the first few sentences? Are your services, prices or price ranges, locations and policies stated plainly and consistently everywhere they appear, including structured data and business profiles?
- Evidence and trust. Are factual claims sourced? Is it clear who wrote the content and how it was produced? Do independent sites mention your brand in the context you want to be known for? Unsupported superlatives are easy for a reader, or a model, to discount.
- Measurement. Have you set up the Search Console generative AI report and Bing Webmaster Tools AI Performance? Have you written down 20 to 50 real buyer questions you want to appear for, so you can sample them in AI assistants every month and compare like with like?
If you can tick all five, you need refinement, not an overhaul. If layer 1 or 2 fails, fix that first. It is usually the cheapest win available.
AI SEO vs GEO: where the line is
The terms overlap, and many agencies use them interchangeably. We separate them like this. AI SEO is the broad discipline. It covers using AI in the SEO process and optimizing the whole site for AI-powered search, with classic SEO as the base. Generative engine optimization (GEO) is the narrower practice of improving how often, and how accurately, generative engines cite and mention your brand. Its tools are prompt tracking, citation share and off-site mention building. "Answer engine optimization" (AEO) is mostly a third name for the same answer-focused work.
GEO has an academic origin. A 2023 paper accepted to KDD 2024 reported that content changes such as adding citations and statistics could boost visibility in generative engine responses "by up to 40%" on the authors' benchmark. It also found that what works varies by domain (Aggarwal et al., arXiv 2311.09735). That is a lab result, not a promise for your site. If citation tracking and brand mentions in ChatGPT, Perplexity and Gemini are your main goal, see our generative engine optimization agency page.
How to choose an AI SEO agency
Selection criteria
- SEO fundamentals first. Ask for their technical audit process. If they can't explain crawling, indexing and rendering clearly, the AI layer won't save you.
- Source-backed method. They should be able to point to Google, Microsoft, OpenAI or Perplexity documentation for what they recommend, and say plainly where they are testing rather than following documented guidance.
- Honest measurement. They should explain what can be measured today (Search Console gen-AI impressions, Bing AI citations, sampled prompts) and what can't.
- Human review of AI output. They should show you how AI-assisted drafts are fact-checked and edited before anything is published.
- Industry fit. Ask whether they understand your buyers. Ecommerce, SaaS and local service businesses get cited for different kinds of questions.
- Transparency. You should get every change logged, access to your own accounts, and no lock-in over content or data.
Questions to ask before you sign
- What will you change in the first 30 days, and why those things first?
- Which of your recommendations are documented by Google or the AI platforms, and which are experiments?
- How do you measure AI visibility, which prompts will you track, and how often will you report?
- How do you use AI in your own workflow, and who reviews the output?
- Do you follow Google Search Essentials, and will you share all site changes with explanations? (Google suggests asking both, in its own hiring checklist.)
- What results do you expect, in what timeframe, and how will we know if it isn't working?
- What do you need from us: developer time, subject experts, approvals?
- Can we see a sample report and a sample audit, with client details removed?
Red flags
- Guaranteed rankings, guaranteed AI Overview placement or guaranteed ChatGPT citations.
- Claims of a "special relationship" with Google or OpenAI, or of "submitting" your site to an LLM index.
- llms.txt or "AI schema" sold as the core of the strategy.
- Plans to publish hundreds of AI-written pages quickly, which risks Google's scaled content abuse policy.
- Reports full of "AI visibility scores" with no explanation of how prompts were chosen or sampled.
- Case studies with big percentages but no dates, baselines or way to verify them.
How Mediardx runs an AI SEO engagement
Mediardx is an SEO agency based in India. We work remotely with businesses in the US, UK, Canada and Australia. Our AI SEO services follow the same rule as this page: do what the primary sources support, label experiments as experiments, and never invent results.
In practice, every engagement starts with the readiness checklist above, run against your site. We fix access and eligibility issues first. Then we rebuild your most important pages so they answer buyer questions clearly and back every claim with a source. In parallel, we set up measurement in Search Console, Bing Webmaster Tools and a fixed prompt set, so changes can be compared month to month. We use AI tools for research, clustering and quality checks. A person reviews everything before it reaches your site, and we tell you where automation was used.
AI SEO rarely stands alone. It usually sits on top of technical SEO, content and authority work, and it adapts to your sector, whether that is ecommerce, SaaS or local services. You can read how we work step by step on our process page and ask for engagement options through our contact page.
What to expect, and when
Timelines depend on your site's starting point, competition and how quickly changes get implemented. As a typical pattern, not a guarantee: access and eligibility fixes can be live within the first few weeks. Rewritten key pages usually follow over the first two to three months. Changes in Search Console impressions and sampled AI answers generally need several months of consistent data before you can read a trend. AI answers also vary from one run to the next, which is why we track a fixed prompt set over time instead of reporting single screenshots.
If an agency promises faster or bigger results than this without explaining how, ask them to show their working.