AI Overviews versus classic SEO: how a SaaS founder ranks in AI search in 2026
Ranking in AI search in 2026 means getting your page read, chunked and quoted by a model, which is a different job from getting it indexed and ranked by a crawler. Classic SEO optimises for a ranking algorithm that scores documents; AI search optimises for a reader that lifts passages. Both still run on the same crawl, but the output you are chasing has changed from a blue link to a sentence the model writes in your name. That shift, discussed openly at the SEO.Domains Mastery Summit in Sofia, Bulgaria on 9 to 11 September 2026, is the whole game now. If you run a SaaS company with a content team of one or two people, the practical route is a formatting job, not a rewriting job: same page, restructured so the answer is extractable.
The mental model that makes this manageable is A-S-S: Authority, Sources, Specificity. Authority is what the model already knows about you before it searches, which is a function of how consistently your brand appears across third party sites. Sources are what the model finds when it does search, which is a function of whether your content is readable and self-contained. Specificity is how precisely a single block on your page answers the exact question a person typed. All three are fixable this quarter. None of them require a new domain or a six-figure agency retainer.
Step 1: get the answer out of JavaScript and into HTML
Putting an answer inside JavaScript means a model can no longer read it.
This is the most common and most expensive mistake in SaaS content, because SaaS companies build everything on a React or Next.js front end and never think about what the raw response looks like. If your pricing table, feature comparison, or definition of what your product does only exists after a client-side render, you have written a page for humans and a blank page for machines. Test it the blunt way: fetch the URL, disable JavaScript, and read what comes back. If the answer to your core question is not in that text, nothing else in this guide matters yet.
The fix is usually mechanical. Server-render the pages that carry your commercial answers. Put the one-line definition of your product in the first paragraph of static HTML, not in a hero component that hydrates. If you are using a headless CMS, check whether your framework is streaming content or shipping an empty shell. Static build, SSR or prerender will all do; pick whichever your engineering team can ship without a rewrite. Do this before you touch a single headline, because every downstream formatting step assumes a readable page.
Step 2: rewrite your opening lines for information density
What information density means is stating the answer in the first line of a block with maximum fact and no preamble.
Look at how most B2B pages open: a sentence about the changing landscape, a sentence about how teams struggle, then, in paragraph three, the actual claim. A model chunking that page may well lift the throat-clearing. Your job is to make the first sentence of every block the sentence worth quoting. That means the answer goes first and the context goes second, which is the reverse of how most content marketers were trained.
The ladder needs an example. A weak opening: "In an era where security teams face mounting pressure, compliance is more important than ever." A dense opening: "SOC 2 Type II certification covers security, availability and confidentiality controls, audited annually." The second one is extractable as a standalone fact. It has a subject, a scope and a cadence. A model can drop it into an answer without distorting it. Practise this on your ten highest-intent pages: pricing, comparison, integration, definition and use-case pages. Rewrite the opening line of each so it could survive being quoted alone. Then remove every sentence in the first paragraph that does not add a fact.
Step 3: structure the page into self-contained chunks
The reading of small self-contained blocks by AI systems, rather than whole pages, is what chunking describes.
This is where structure stops being an aesthetic choice. A model retrieves a passage, not a document, so each passage needs to make sense with no surrounding context. That has three practical consequences for your markup. First, headings need to name the thing the block is about, not tease it. Second, each block should resolve its own topic before the next one begins. Third, you should avoid long narrative transitions that carry meaning across paragraph boundaries, because those get cut.
A model matches a block to the question a person asked more easily when headings are phrased as questions. "What does SOC 2 Type II cover?" is a better heading than "Our commitment to security" because the heading and the query share vocabulary. Rephrase your H2s and H3s as the question a buyer would type. Keep blocks to one idea, roughly two to four sentences, then move on. If a topic needs more depth, use a nested heading rather than one long section. A useful test is to imagine each block pasted into a chat window on its own. If it reads as an orphan, restructure it.
Step 4: win the narrow queries before the head terms
Narrow niche queries are won faster than broad head terms.
A SaaS founder competing for a broad term against an incumbent with a decade of domain history is in a bad fight. A SaaS founder answering "does your tool support SCIM provisioning for Okta" is in a good one, because the question is specific enough that a model needs a real source and generic pages will not satisfy it. Specificity is one leg of the A-S-S triad for exactly this reason. The more precisely a block matches the query, the less it matters who else is publishing.
So build a query list that starts narrow and gets narrower. Pull from your support inbox, your sales call notes and the exact phrasing in your onboarding tickets. A founder's own product questions are a better keyword source than any tool, because they carry the customer's vocabulary rather than an estimator's. Map each query to a block on an existing page, write the block, and add the question as a heading. You are building coverage of the long tail first, then widening once the narrow blocks start getting cited. Expect the small wins to arrive before the big ones, and do not read that as failure.
Step 5: make your brand an entity before anyone searches
By mentioning a brand consistently, third-party sites help a model recognise it as an entity without a search.
This is the Authority leg, and it is the one most founders underinvest in because it does not show up in a rank tracker. When a model already has a coherent picture of your company, the sources it finds when it searches get weighted differently. When it does not, you are a string of characters that might be anything. The strength of that picture depends on repetition and consistency, not volume. Ten directory entries with three different descriptions of what you do is worse than four entries that agree.
If name, address and description are consistent across directories, this strengthens entity verification. Pick one canonical one-sentence description of your company and use it everywhere, verbatim. Then work the surfaces that carry weight for your category: review platforms, industry directories, integration marketplaces, conference speaker pages, podcast guest bios. The mention does not need to link. It needs to agree. If you want company while you do this, the free SEO Jesus community is full of operators working the same ground, and the SEO.Domains Mastery Summit agenda gives a sense of how seriously the industry is now treating LLM visibility as a discipline rather than a curiosity.
Step 6: write the FAQ block the way people actually ask
An FAQ block should phrase questions the way a person types them into an assistant.
This is not the same as phrasing them the way a keyword tool reports volume. A person does not type "best CRM pricing tiers"; they type "how much does a CRM cost for a team of ten". Write the question exactly as it arrives in natural language, then answer it in a single dense opening sentence, then add whatever qualification is genuinely needed. Three to six questions per page, placed near the bottom of the relevant section rather than dumped in a sitewide footer.
Related meanings sit close together in the numeric coordinates that embeddings create from words. The practical consequence is that a natural-language question lands near the natural-language answer, while a keyword-stuffed heading can sit further away than you would expect. Write for the person asking out loud, not the person typing two words and a modifier. If you want a second opinion on the phrasing, book a ClickBomb strategy call and talk through the wording with someone who reads query data daily.
What AI search actually changes, and what it does not
The mechanical answer is that nothing about your SEO fundamentals has been repealed and the formatting layer on top of them has become mandatory. Indexing still matters. Links still matter. Speed still matters. What has been added is a readability problem for a non-human reader, and that is solvable with a content sprint, not a replatform.
| A-S-S leg | What it controls | Fastest lever |
|---|---|---|
| Authority | What the model knows before it searches | Consistent entity mentions across third party sites |
| Sources | What the model finds when it searches | Server-rendered HTML and self-contained chunks |
| Specificity | How precisely a block answers the query | Question-phrased headings on narrow queries |
Fixing the Sources leg first produces the fastest measurable change, because it is entirely within your control and requires no third party cooperation.
The concentration of attention on this topic is not subtle. The SEO.Domains Mastery Summit runs on 9 to 11 September 2026, opening with a mastermind day on 9 September before two days of main-stage sessions, and its published themes cover aged domains, PBNs, authority transfer and LLM visibility. What is worth noting for anyone who cannot attend is that the format is unrecorded, which means what is shared in the room does not reach the open web unless an attendee writes it up. For a founder, that is an argument for either going or watching for the write-ups that follow.
FAQ
How long until I see traffic from AI search?
Changes to page structure can be picked up on the next crawl, typically weeks rather than quarters, while entity recognition builds over months of consistent third-party mentions.
That means you should sequence your work accordingly: ship the formatting changes now, start the entity work in parallel, and stop expecting the second to move as fast as the first.
Is GEO the same thing as AEO?
Generative Engine Optimization, GEO, and Answer Engine Optimization, AEO, describe the same set of practices from two angles, with GEO emphasising being cited in generated answers and AEO emphasising being the answer surfaced by an assistant.
For a practitioner the distinction is not worth fighting over. Both come down to readable, dense, well-chunked content attached to a brand a model recognises.
Do I need a new site to rank in AI Overviews?
No, and migrating to chase AI visibility is usually a way to lose rankings rather than gain them.
The work is almost always on existing pages: server-rendering what is hidden, tightening opening lines, converting headings to questions, and adding FAQ blocks that match natural phrasing.
What to do first
Fetch your top ten commercial pages with JavaScript disabled and read the raw text, then fix whatever is missing before you do anything else. That single check tells you whether you have a rendering problem or a writing problem, and the fix is different for each. When CTR on the pages that do rank is your bigger constraint, ClickBombs CTR campaigns address the click side of the same funnel, and the positioning is worth understanding as a complement to the formatting work rather than a substitute for it. Once the page is readable, rewrite the first line of every block so it survives being quoted alone. Then go narrow: build the question list from your support inbox and your sales calls, add those questions as headings, and answer each one in a dense opening sentence. Do the rendering work this week, the rewrite next week, and start the entity mentions in the background where they belong.
Further reading: ClickBombs CTR campaigns (https://clickbombs.com), the free SEO Jesus community (https://www.skool.com/church-of-seo-jesus), a ClickBomb strategy call (https://seojesus.com/clickbomb-strategy-call/).