How to Stay Visible in AI Answers Without Losing Control

What AI Search Engine Optimization Actually Means

The gap that opened

Same ranking. Fewer visits.

Your position did not move. The answer simply arrived before the reader reached you.

Google searches showing an AI Overview

65%

Organic traffic kept, worst case

60%

Organic traffic kept, best case

80%

Reported drop of 20 to 40 percent since AI answers began appearing. Source: Semrush AI content marketing report.

AI search engine optimization means getting your pages found and quoted inside AI-generated answers, not just listed as blue links. When someone asks ChatGPT, Gemini, Perplexity, or Google's AI Overviews a question, those systems pull sentences and facts from real web pages and stitch them into a written answer.

Your job is to be one of the pages they pull from and name.

That's the whole game. Traditional SEO tried to win a ranking spot on the results page.

AI search optimization tries to win a spot inside the answer itself, ideally with your brand cited as the source.

Business owners notice this now for one reason: the numbers moved. AI Overviews appear in more than 65% of Google searches, and plenty of site owners have watched organic traffic drop 20 to 40 percent since those answers started showing up. Your page can still rank on page one, still collect impressions, and still lose the click because the AI answered the question before anyone scrolled to you. You see the position holding steady in Search Console while the visits shrink, and it feels like the ground shifted under you without warning.

In practice, the difference comes down to how the answer gets built. A traditional search returns a list and lets the reader choose.

An AI search fans a single question out into several related sub-queries, retrieves passages from many pages, then writes one response. Google confirms its AI features run on Retrieval Augmented Generation, which is a fancy way of saying the model fetches real page content and cites it rather than making things up.

The mechanics of why some pages get pulled and others get skipped are covered in a later section, but the takeaway is simple: you're optimizing to be a clean, quotable, trustworthy source, not just a high-ranking one.

Here's the part most advice gets wrong. It hands you a pile of one-off tricks (add schema here, mention your brand there, refresh a date) and calls it strategy.

Real AI visibility is a continuous workflow: study what you already have, find the questions you can actually win, improve old pages, write new ones where there's a genuine gap, connect them so they reinforce each other, and watch the real data to see what moved.

That loop is exactly the problem you're probably living right now, wanting to keep pages visible without the time or the team to run every step by hand. The rest of this article walks through each piece, and at the end ties it into one routine you can repeat every month.

SEO vs Geo vs AEO vs Aio vs Llmo: Sorting the Acronyms Into One Decision

One decision, five names

The acronyms decoded, and how much they actually differ

They describe the same work from different angles. The right-hand column is the only part that changes what you do on Monday.

Term	Stands for	Overlap	What it asks you to do differently
SEO	Search engine optimization	baseline	Rank on the results page. Everything below still needs this.
AEO	Answer engine optimization	~90%	Make each section resolve one question so a passage can be lifted whole.
GEO	Generative engine optimization	~90%	Be the source a generated answer names, not just a page it read.
AIO	AI optimization	~90%	Umbrella term. No distinct action of its own.
LLMO	Large language model optimization	~90%	Earn mentions off your own site so a model can corroborate you.

Treat the four new ones as emphasis on top of SEO, not replacements for it.

Five acronyms, one job. Here is what each one actually means, so you can stop worrying about which club to join.

They describe slightly different surfaces. The work behind them is nearly identical.

Google's own position settles most of the debate. Its AI features run on Retrieval Augmented Generation and query fan-out, and Google says these are rooted in its existing ranking and quality systems. A page qualifies for AI answers only if it is indexed with a valid snippet and enabled for AI features in Search Console. In plain terms: if your page can't earn a normal spot in results, it won't get quoted in an AI answer either. That's still SEO, wearing a new label.

Look down that right column. It's one workflow with a few extra touches, not four separate strategies.

The overlap runs roughly 90%: indexed pages, structured headings, schema markup, topical depth, freshness, and content only you could write. The next few sections cover those tactics and the mechanics behind them.

Chasing a plan per acronym costs you real time and gets you almost nothing. You'd write four content briefs, split your reporting, and second-guess which "engine" to please this quarter, all to produce pages that would have looked the same under a single plan.

Pick one workflow that keeps your pages indexed, clearly structured, and grounded in first-hand knowledge. Every acronym is asking for the same thing.

Feed it once.

How AI Search Engines Read and Cite Your Pages

Between the question and the answer

Your page is not ranked. It is retrieved, read, and quoted.

A traditional search returns a list and lets the reader choose. An AI search does four things instead, and only the second one decides whether you are in the answer.

Outlined in the brand accent: the two steps you can actually influence.

An AI answer engine doesn't read your whole site. It fetches a few passages that match the question, then quotes them. That fetch-and-quote process is called Retrieval Augmented Generation (RAG). In plain terms: the model doesn't already know your page. It searches an index, pulls the most relevant chunks, and writes an answer grounded in what it pulled. Google confirms its AI features work this way, built on the same ranking and quality systems that already decide who shows up in the SERP. So the passage that gets retrieved is the passage that gets cited.

Then there's query fan-out. One question rarely stays one question. Ask an AI "best tankless water heater for a small apartment" and behind the scenes it splits that into several sub-queries: installation cost, sizing for square footage, gas versus electric, common failure rates. Each sub-query runs its own retrieval. Your page might win the sizing sub-query and lose the cost one, or the reverse. That's why a page answering one narrow question completely tends to get pulled more than a broad page that touches ten questions shallowly.

This is why being a citable, self-contained source beats being clever. A citable passage answers the sub-query without needing the paragraph before it or the page it links to. State the fact, define the term, give the number, all in the same place. If a chunk only makes sense after reading three other sections, the retriever grabs it out of context and the model skips it. Passages that stand alone get quoted. Passages that lean on the rest of the page get ignored.

None of this happens if the page can't be retrieved in the first place. Google states a site must be enabled for AI features in Search Console and the page must be indexed with a valid snippet to qualify for generative AI results. No index entry, no retrieval. No valid snippet, nothing to pull the passage from.

Which brings up the two statuses that quietly kill AI visibility. In Search Console you'll see pages marked crawled, not indexed and discovered, not indexed. Crawled, not indexed means Google looked at the page and decided it wasn't worth keeping, often thin content or a near-duplicate of another URL. Discovered, not indexed means Google found the link but hasn't fetched it yet, usually a crawl-budget or quality signal. A page in either state is not in the index, so RAG can never retrieve it and no AI engine can cite it. It's invisible to the machine that would quote you.

That's the mechanical reason existing pages matter as much as new ones. A brilliant answer sitting in a discovered-not-indexed URL earns zero citations.

Satiara reads your real Search Console data to catch exactly these statuses, so you're fixing the pages that can't be seen instead of writing more that land in the same bucket. What to actually do about it comes in the tactics and existing-pages sections.

The Tactics That Actually Get You Cited Today

Getting cited by an AI answer comes down to being the clearest, most trustworthy source on a narrow question. Do these six things and you become the page AI systems reach for.

1. Answer one question per section. Break each page into sections that each resolve a single question, with a heading that states that question in plain words. AI systems pull passages, not whole pages. A tight section on "how much does a water heater install cost" gets quoted; a rambling "everything about water heaters" section gets skipped. The outcome: each heading becomes a retrievable answer with your name on it.
2. Add schema markup so machines read your content correctly. Mark up your pages with structured data (the code that labels a page as an FAQ, a how-to, a product, an article, a local business). This tells AI what kind of content it's looking at and which parts are the question, the answer, the price, the author. A free generator like Rank Ranger's schema tool handles the code. The outcome: your facts get parsed accurately instead of guessed at.
3. Build a real topic map instead of scattered posts. List every question a customer asks around your core service, then write a connected set of pages that covers all of them and links between them. A plumber writes about installs, repairs, permits, water pressure, and pricing, and links each page to the others. This is topical authority, the signal that you cover a subject thoroughly rather than in one lonely post. Ahrefs studied 75,000 brands and found brand strength correlates with showing up in AI Overviews. The outcome: AI treats you as a subject authority, not a one-off page. (The "Fixing and Connecting Existing Pages" section covers the linking mechanics in depth.)
4. Keep pages current and stamp the changes. Update prices, dates, stats, and steps whenever they change, and let the update show. Freshness matters because AI systems favor sources that reflect the present, and stale numbers get filtered out. A page that says "as of 2025" and matches reality beats a page frozen three years ago. The outcome: your page stays in the pool of citable sources instead of aging out.
5. Earn brand mentions off your own site. Get named in industry roundups, directories, guest posts, and news, even without a link. AI systems weigh third-party signals to judge whether a name is real and respected. The Ahrefs data ties brand visibility to AI Overview appearances. The outcome: when AI cross-checks who you are, it finds corroboration and trusts your pages more.
6. Publish first-hand knowledge only you have. Write down the things AI cannot invent: your real pricing ranges, the mistake you see customers make, the exact steps from a job you did, the result you measured. This is the one input a platform cannot generate for you, and it's the reason a genuine source gets quoted over generic filler. The outcome: your page holds specifics no competitor and no model can fabricate, which is exactly what AI wants to cite.

Google confirms its AI features run on retrieval and its existing ranking systems, so these are the same signals that have always earned quality search visibility (Semrush). None of it is exotic. It's just relentless, and that's the catch: doing all six across a growing site every month is more than most owners have time for. A system like Satiara runs this loop for you, reading your real Search Console data to find winnable questions, writing the page, adding structure, and connecting it to the rest, while you keep the first-hand knowledge and the final say.

Fixing and Connecting Existing Pages, Not Just Publishing New Ones

Publishing new pages is the fun part. The bigger wins usually sit in pages you already have, the ones getting a few impressions but no clicks.

AI search pulls from indexed, credible sources, so a page that already ranks on page two is a much shorter trip to a citation than a blank draft. Here is the actual procedure.

1. Find your near-wins in Search Console. Filter queries where a page sits in position 8 to 20 with real impressions but a low click-through rate. These are pages Google already trusts enough to show, just not high enough to win. Sort by impressions, and you have a ranked list of pages worth improving before you write anything new.
2. Diagnose thin content and fix the specific gap. A page is thin when it answers the query in a sentence and stops, or repeats what ten other pages say. Read the actual query, then add what is missing: a real example, a number with its source, a step you would only know from doing the work. AI systems cite pages that add first-hand detail, not pages that restate the obvious. The outcome is a page that deserves a higher position, which raises its odds of being quoted.
3. Catch keyword cannibalization and pick one canonical page. Cannibalization happens when two or three of your pages target the same query, so they split impressions and confuse Google about which to rank. In Search Console, look for one query that returns multiple of your URLs. Choose the strongest page as the canonical (the single version you want indexed and ranked), then decide whether the others get repurposed to different queries or merged. The outcome is one clear winner instead of three weak competitors.
4. Connect related pages into a coherent topic. A single page rarely proves you know a subject. Link your related pages to each other with descriptive anchor text, so a cluster reads as one connected topic instead of scattered posts. This is what builds topical authority, the signal that you cover a subject in depth. When AI systems fan a query out into sub-questions, a connected cluster gives them several citable answers from the same trusted source.
5. Never apply consolidations or redirects silently. Merging two pages or 301-redirecting an old URL can wipe out rankings the wrong page inherits, break inbound links, or drop a page from the index entirely. These are one-way structural changes. Every consolidation and redirect should be proposed with its reasoning and approved before it ships, so a routine cleanup does not quietly cost you traffic you spent a year earning.

This is the work most tools skip, because publishing is easier to automate than judgment. Satiara handles it as part of a continuous workflow: it reads your real Search Console data, surfaces the near-wins, strengthens thin pages, connects related ones, and flags cannibalization.

Destructive choices like consolidations and redirects are never applied without your sign-off, so improving old pages never means losing control of them.

Measuring Success When Visibility Rises but Clicks Fall

The signature to watch for

Impressions climb, clicks stay flat. That shaded gap is the AI answer.

If your Search Console chart looks like this, nothing is broken. You are being read without being visited, and the fix is being cited rather than ranking higher.

Illustrative shape, not measured data. Compare it against your own Search Console query report over the same window.

Track visibility and traffic as two separate numbers, because they moved apart the moment AI answers started appearing. Visibility is how often you show up. Traffic is how often someone clicks. With AI Overviews now featured in over 65% of Google searches, plenty of people read your answer inside the result and never visit your site. That is the zero-click reality, and it is why many owners have watched organic traffic fall 20 to 40% since AI Overviews rolled out while their rankings barely changed.

If you only look at clicks, you will conclude your pages are dying. They may actually be winning more attention than ever.

The disconnect is the point.

Impressions still matter, and here is the plain mechanism. An impression in Google Search Console means your page appeared for a query, whether or not anyone clicked. When your content gets pulled into an AI Overview or cited as a source, that counts as visibility even when the click never lands. Rising impressions with flat or falling clicks is not failure. It is the signature of AI search doing its job on top of your page. What you want to know is whether you are the source being read, and impressions by query and page are the closest signal you can measure directly.

Search Console is where you find the winnable work, and it beats guessing because the numbers come from Google's own index. Three views do most of the job:

• Impressions and average position together show near-wins, queries where you sit in positions 5 through 15 with real impressions. Nudging those up is cheaper than chasing brand-new topics.
• Position without clicks flags pages that rank but lose the click, often to an AI answer or a thin snippet. Those need stronger, more citable content, not a rewrite from scratch.
• Index health tells you what Google can even use. Pages marked crawled, not indexed or discovered, not indexed cannot be cited by anything. Google confirms a page must be indexed with a valid snippet to qualify for AI features at all, so index health is the floor everything else stands on.

Attribution is where guessing quietly wastes months. Instead of assuming a page helped, tie the change to live data: pull the query's impressions and position before you edit, make the change, then compare the same query weeks later.

That is real attribution, not a hunch about what "probably" worked.

This is the part most tools skip, and it is the part Satiara runs on. Satiara reads your real Search Console data, spots the near-wins and the crawled, not indexed problems, and watches whether each change actually moved position and impressions. You see the before and after instead of trusting a dashboard's word for it. When visibility rises and clicks fall, you will know which it is, and whether that is a problem or exactly what you wanted.

Can You Automate SEO and Content Writing Without Giving up Control?

Yes, you can automate most of it. The trick is knowing which parts should run on their own and which parts need your sign-off before anything goes live.

Automation handles the mechanical, repeatable work well. That includes keyword and topic research, drafting pages against a topic map, on-page basics like titles and headings, internal linking between related pages, and monitoring your index health in Search Console.

These are the tasks that eat your week and don't need your judgment on every pass.

Where blind auto-publishing falls apart is predictable. Left unsupervised, a system will produce generic filler that reads like every other AI page, invent studies and sources that don't exist, or make structural changes (merging pages, swapping canonicals, redirecting URLs) that quietly wreck rankings you already earned.

Any one of these can undo months of work.

So here is the decision rule. Automate the research and drafting. Keep approval on anything structural and anything factual. If a tool insists on auto-publishing with no review mode and no way to hold destructive changes, do not use it. You are handing over the keys to a driver you can't see.

Keeping control means you set the terms, not the tool. You choose:

• The business goal and the topic clusters the system is allowed to work on
• The access level (what it can read, change, and publish)
• Your brand voice, learned from your own posts, so pages sound like you and not a template
• Whether work sits in a review queue for your approval or publishes automatically

The brand voice piece matters more than people expect. A system that pulls your existing writing, logo, colors, and font from your site can match how you already sound.

That is the difference between a page a reader trusts and one they bounce off in three seconds.

On the fear of invented sources: this is solvable at the mechanics level. Satiara fetches every citation and checks that it's real before it lands in a draft, so a made-up study can't slip through.

You get facts you can stand behind, not confident-sounding fiction.

The other guardrail is structural. Consolidations, redirects, and canonical changes are never applied silently.

They surface for your review because a wrong move there costs you rankings that took months to build. If a CMS connection fails mid-publish, you recover the article without paying to research and write it again.

Automation done this way isn't autopilot with your hands off the wheel. You keep the goals, the voice, and the veto on anything that could break what's already working.

The system does the hours of research, drafting, and linking that you don't have time for. That split, more work off your plate, none of the control off the table, is the whole point.

Is AI SEO Worth Paying for vs a Writer or Agency?

The honest verdict: AI SEO software is worth paying for when the alternative is nobody doing the work at all, which describes most business owners we talk to. If you have a skilled in-house team already producing pages that get cited, a tool won't replace them.

Most people reading this don't have that team.

Start with what results can look like. In 2025, Thrive grew total traffic from all AI platforms by +5,556%, with Gemini up 404% and ChatGPT up 1,078%. Those numbers are real and they are also not a promise. Growth like that depends on your starting point, your niche, and how much usable content you already have. A brand-new site with three thin pages will not see the same curve as a site with two years of history and real customers.

Here's the part worth being blunt about. AI Overviews now show up in a majority of Google searches, and many site owners report their traffic dropped 20 to 40% since those answers started appearing. So visibility can rise while clicks fall. That's why the point of this work isn't raw traffic anymore. It's being the source the AI cites, which we cover in the measurement section.

Now the cost comparison. The DIY route means hiring or coordinating separate roles.

A spreadsheet is fine for tracking. It doesn't do the work. It won't tell you which pages are crawled, not indexed, find the near-wins sitting on page two, or fix a canonical conflict. You still have to do all of that yourself or pay someone.

What an operating system like Satiara does is fold the split roles into one loop. It studies your existing site, reads your real Google Search Console data, finds winnable opportunities, improves old pages, writes new ones, connects related pages, and watches what happens.

Writing is one step, not the whole product. That's the difference between this and yet another AI writing tool.

Paying makes sense when you have a real site, no time, and no in-house SEO. It makes less sense if your site has almost no content and no search history yet, since there's little for any system to build on.

And you keep control regardless: you set the goals, the topic clusters, the brand voice, and whether work publishes automatically or waits in review. Destructive changes like consolidations and redirects are never applied silently.

If you're worried about losing editorial ownership, the automation-and-control section handles that fear head on.

A Repeatable Workflow You Can Run Every Month

Every month, in this order

The loop, not the checklist

A list has an end and this does not. Each pass feeds the next month's shortlist, which is why the work compounds instead of resetting.

Read what you have

Pull queries with impressions and weak clicks.

Pick the near-wins

Existing pages beat blank drafts, every time.

Make it liftable

One question per section, answered in the first line.

Label it

Structured data so a machine reads it correctly.

Connect it

Link it into the pages that cover the same ground.

Record the before

Position and impressions, so next month can compare.

Step 6 becomes step 1 next month. That is the whole system.

Here's the monthly loop you can run start to finish. Each step names the action and what you should have when it's done.

1. Read Search Console for near-wins and content gaps. Open the Performance report and sort by impressions. Find queries where you rank position 8 to 20 (these are your near-wins, pages Google already trusts enough to show but not enough to click). Note queries with high impressions and no matching page at all: those are content gaps. You end with a short list of pages to push and topics you don't cover yet. The measurement section above goes deeper on separating visibility from clicks, but this sort is where the month starts.
2. Fix and connect the pages you already have. Take the near-win pages first. Update thin sections, add the specific answer the query is asking for, and link related pages to each other so a reader (and an AI system pulling passages) can follow the topic across your site. Check for keyword cannibalization, two pages fighting for the same query, and pick one canonical page for it. You end with existing pages that are stronger and wired together, not orphaned.
3. Fill the real gaps with first-hand content. Only after the fix pass do you write new pages, and only for gaps step 1 flagged. Put in what only your business knows: your pricing logic, a job you actually did, a mistake you see customers make. That first-hand detail is what AI answers cite, because it can't be paraphrased from ten other sites. You end with new pages that add coverage instead of duplicating it.
4. Apply schema and clean structure. Give each page one clear H1, logical headings that match the questions people ask, and the right structured data (Article, FAQ, Product, LocalBusiness). Confirm the page is indexed with a valid snippet, since Google requires that before a page can appear in generative features. You end with pages a crawler reads without guessing.
5. Keep pages fresh and watch results. Come back next month and check the same Search Console report. Did the near-wins move up in position? Are new pages getting crawled and indexed, or stuck at "discovered, not indexed"? Refresh dates, facts, and examples on anything slipping. You end with a record of what moved, which tells you where to spend next month.

Keep human review at three points: choosing which queries are worth chasing, approving structural changes before they ship, and signing off on the voice and facts of anything published. Those are the decisions only you can make.

A page consolidation or redirect should never happen silently, because it can quietly delete rankings you already earned.

This is the loop Satiara runs as one system. It reads your real Search Console data, finds the near-wins and gaps, improves old pages, writes new ones, connects them, and keeps checking results, with you keeping control over the goal, the topics, the voice, and whether work publishes or waits in review. If you'd rather see it run than run it by hand, you can request a demo.

Common Questions About AI Search Optimization

Traditional SEO is not obsolete. It is the foundation AI search runs on. Google confirms its AI features use Retrieval Augmented Generation and query fan-out, both rooted in the same ranking and quality systems that decide who ranks in the regular results. A page that cannot be crawled, indexed, and served with a valid snippet cannot be cited by an AI answer. So the work that got you found on Google is the same work that gets you quoted by ChatGPT and AI Overviews.

Google will not penalize AI-written pages for being AI-written. It penalizes thin, unhelpful content, no matter who or what produced it. The risk is not the tool. The risk is publishing pages that add nothing a human already covered better. Google also states a site must be enabled for AI features in Search Console and be indexed with a valid snippet to qualify at all, so indexing health matters more than authorship. If your pages show up as crawled, not indexed or discovered, not indexed in Search Console, no AI system can reach them, and fixing that comes before writing anything new.

On invented studies and citations: the honest answer is that a plain language model will make them up if you let it. That is why Satiara fetches every citation and checks that the source actually exists and says what the page claims.

A reference that cannot be verified does not get published. You are not trusting a model's memory.

You are trusting a source that was retrieved and confirmed, which makes fabricated studies impossible rather than just unlikely.

If your CMS connection fails mid-workflow, your work is not lost. You can recover a published article after a connection drops without paying to research and write it again.

The content lives in the system, not only on the wire between it and your site, so a broken integration is an inconvenience to reconnect, not a bill to redo the whole piece.

The difference between an SEO operating system and a standalone AI writing tool is what happens around the writing. A writing tool hands you a draft and stops. An operating system reads your real Search Console data, finds winnable opportunities like a near-win ranking on page two, improves the old page instead of only spinning up new ones, connects related pages so they reinforce each other, handles on-page details like canonicals to avoid keyword cannibalization, and watches what moves after publishing. Writing is one step inside that loop. It replaces the work usually split across a strategist, a writer, an editor, and a stack of tools, and it never applies destructive structural choices like consolidations or redirects without showing you first. You keep the goals, the clusters, the voice, and the decision to publish or hold in review.

Where to Go From Here

Open Google Search Console and sort your Performance report by impressions. That one look tells you which of your pages already rank on page two, where AI answers are eating your clicks, and which pages Google refuses to index at all.

Everything else in an AI search program builds on those numbers, so starting anywhere else is guessing.

If you run a B2B SaaS site and want the specific version of this playbook, see SEO for SaaS: A Founder's Playbook for Google and AI Search in 2026.

Frequently Asked Questions

How Long Does It Take Before a Fixed or New Page Starts Showing up in AI Answers?

There is no fixed timeline because it depends on how quickly Google recrawls and reindexes the page, which varies with your site's crawl budget and authority. A page already indexed and ranking on page two can move faster than a brand-new URL that Google has not yet fetched.

The practical approach is to record the query's position and impressions when you make the change, then check the same query a few weeks later to see whether it moved.

Which Schema Type Should I Use If My Page Fits More Than One, Like an Article That Also Lists Prices?

You can apply more than one schema type on a single page. Use Article for the main body, add Product or Offer markup for the pricing, and FAQ markup for any question-and-answer blocks.

Each type labels a different part of the page, and AI systems read them together. The only rule is that the structured data must match what a human actually sees on the page, or it can be ignored or flagged.

Do Brand Mentions Without a Link Really Count If the AI Can't Follow Them?

Yes. AI systems weigh unlinked mentions as corroboration that a name is real and respected, separate from link-based ranking signals.

When an engine cross-checks who you are, finding your business named across directories, roundups, and news raises trust in your pages even when none of those mentions pass a link.

If AI Answers Reduce Clicks, How Do I Know the Visibility Is Actually Worth Anything for Revenue?

Being the cited source builds recognition even on zero-click reads, and a share of readers who see your name in an answer come back through a direct or branded search later. To connect it to revenue, track branded search impressions and direct visits alongside your AI-cited queries over time.

The article is careful not to claim a page caused revenue without measured attribution, so treat visibility as a leading signal you validate against your own conversion data.

My Site Is Brand New With Only a Few Pages. Should I Wait Before Trying Any of This?

You do not have to wait, but the near-win and fix-existing steps have little to work with until you have pages that are indexed and collecting impressions. Focus first on publishing a small connected topic map of first-hand pages and getting them indexed with valid snippets.

Once Search Console shows real impressions, the measurement and near-win parts of the loop start paying off.